Transformer系画像認識モデル論文リンク集

趣旨

Attentionを主要なコンポーネントとして構成される画像認識モデルの論文へのリンクを、arXiv.orgへの論文登録日時の順で並べて掲載。公式実装が公開されている場合は、そのリンクも掲載。

関連記事

2020

Visual Transformers

Paper	Visual Transformers: Token-based Image Representation and Processing for Computer Vision
Date	5 Jun 2020
Code
Authors	Bichen Wu, Chenfeng Xu, Xiaoliang Dai, Alvin Wan, Peizhao Zhang, Zhicheng Yan, Masayoshi Tomizuka, Joseph Gonzalez, Kurt Keutzer, Peter Vajda
Org.	Facebook Reality Labs, Facebook AI, UC Berkeley
Abstract	Computer vision has achieved remarkable success by (a) representing images as uniformly-arranged pixel arrays and (b) convolving highly-localized features. However, convolutions treat all image pixels equally regardless of importance; explicitly model all concepts across all images, regardless of content; and struggle to relate spatially-distant concepts. In this work, we challenge this paradigm by (a) representing images as semantic visual tokens and (b) running transformers to densely model token relationships. Critically, our Visual Transformer operates in a semantic token space, judiciously attending to different image parts based on context. This is in sharp contrast to pixel-space transformers that require orders-of-magnitude more compute. Using an advanced training recipe, our VTs significantly outperform their convolutional counterparts, raising ResNet accuracy on ImageNet top-1 by 4.6 to 7 points while using fewer FLOPs and parameters. For semantic segmentation on LIP and COCO-stuff, VT-based feature pyramid networks (FPN) achieve 0.35 points higher mIoU while reducing the FPN module's FLOPs by 6.5x.

ViT(Vision Transformer)

Paper	An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
Date	22 Oct 2020
Code	https://github.com/google-research/vision_transformer
Authors	Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby
Org.	Google Research, Brain Team
Abstract	While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train.

DeiT(Data-efficient Image Transformers)

Paper	Training data-efficient image transformers & distillation through attention
Date	23 Dec 2020
Code	https://github.com/facebookresearch/deit
Authors	Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, Hervé Jégou
Org.	Facebook AI, Sorbonne University
Abstract	Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. However, these visual transformers are pre-trained with hundreds of millions of images using an expensive infrastructure, thereby limiting their adoption.In this work, we produce a competitive convolution-free transformer by training on Imagenet only. We train them on a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop evaluation) on ImageNet with no external data.More importantly, we introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention. We show the interest of this token-based distillation, especially when using a convnet as a teacher. This leads us to report results competitive with convnets for both Imagenet (where we obtain up to 85.2% accuracy) and when transferring to other tasks. We share our code and models.

BoTNet(Bottleneck Transformers)

Paper	Bottleneck Transformers for Visual Recognition
Date	7 Jan 2021
Code
Authors	Aravind Srinivas, Tsung-Yi Lin, Niki Parmar, Jonathon Shlens, Pieter Abbeel, Ashish Vaswani
Org.	UC Berkeley, Google Research
Abstract	We present BoTNet, a conceptually simple yet powerful backbone architecture that incorporates self-attention for multiple computer vision tasks including image classification, object detection and instance segmentation. By just replacing the spatial convolutions with global self-attention in the final three bottleneck blocks of a ResNet and no other changes, our approach improves upon the baselines significantly on instance segmentation and object detection while also reducing the parameters, with minimal overhead in latency. Through the design of BoTNet, we also point out how ResNet bottleneck blocks with self-attention can be viewed as Transformer blocks. Without any bells and whistles, BoTNet achieves 44.4% Mask AP and 49.7% Box AP on the COCO Instance Segmentation benchmark using the Mask R-CNN framework; surpassing the previous best published single model and single scale results of ResNeSt evaluated on the COCO validation set. Finally, we present a simple adaptation of the BoTNet design for image classification, resulting in models that achieve a strong performance of 84.7% top-1 accuracy on the ImageNet benchmark while being up to 2.33x faster in compute time than the popular EfficientNet models on TPU-v3 hardware. We hope our simple and effective approach will serve as a strong baseline for future research in self-attention models for vision.

2021 Q1

T2T-ViT(Tokens-to-Token ViT)

Paper	Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNet
Date	28 Jan 2021
Code	https://github.com/yitu-opensource/T2T-ViT
Authors	Li Yuan, Yunpeng Chen, Tao Wang, Weihao Yu, Yujun Shi, Zihang Jiang, Francis EH Tay, Jiashi Feng, Shuicheng Yan
Org.	National University of Singapore, YITU Technology
Abstract	Transformers, which are popular for language modeling, have been explored for solving vision tasks recently, \eg, the Vision Transformer (ViT) for image classification. The ViT model splits each image into a sequence of tokens with fixed length and then applies multiple Transformer layers to model their global relation for classification. However, ViT achieves inferior performance to CNNs when trained from scratch on a midsize dataset like ImageNet. We find it is because: 1) the simple tokenization of input images fails to model the important local structure such as edges and lines among neighboring pixels, leading to low training sample efficiency; 2) the redundant attention backbone design of ViT leads to limited feature richness for fixed computation budgets and limited training samples. To overcome such limitations, we propose a new Tokens-To-Token Vision Transformer (T2T-ViT), which incorporates 1) a layer-wise Tokens-to-Token (T2T) transformation to progressively structurize the image to tokens by recursively aggregating neighboring Tokens into one Token (Tokens-to-Token), such that local structure represented by surrounding tokens can be modeled and tokens length can be reduced; 2) an efficient backbone with a deep-narrow structure for vision transformer motivated by CNN architecture design after empirical study. Notably, T2T-ViT reduces the parameter count and MACs of vanilla ViT by half, while achieving more than 3.0% improvement when trained from scratch on ImageNet. It also outperforms ResNets and achieves comparable performance with MobileNets by directly training on ImageNet. For example, T2T-ViT with comparable size to ResNet50 (21.5M parameters) can achieve 83.3% top1 accuracy in image resolution 384×384 on ImageNet.

CPVT(Conditional Position encoding Vision Transformer)

Paper	Conditional Positional Encodings for Vision Transformers
Date	22 Feb 2021
Code	https://github.com/Meituan-AutoML/CPVT
Authors	Xiangxiang Chu, Zhi Tian, Bo Zhang, Xinlong Wang, Xiaolin Wei, Huaxia Xia, Chunhua Shen
Org.	Meituan Inc., The University of Adelaide
Abstract	We propose a conditional positional encoding (CPE) scheme for vision Transformers. Unlike previous fixed or learnable positional encodings, which are pre-defined and independent of input tokens, CPE is dynamically generated and conditioned on the local neighborhood of the input tokens. As a result, CPE can easily generalize to the input sequences that are longer than what the model has ever seen during training. Besides, CPE can keep the desired translation-invariance in the image classification task, resulting in improved classification accuracy. CPE can be effortlessly implemented with a simple Position Encoding Generator (PEG), and it can be seamlessly incorporated into the current Transformer framework. Built on PEG, we present Conditional Position encoding Vision Transformer (CPVT). We demonstrate that CPVT has visually similar attention maps compared to those with learned positional encodings. Benefit from the conditional positional encoding scheme, we obtain state-of-the-art results on the ImageNet classification task compared with vision Transformers to date.

PVT(Pyramid Vision Transformer)

Paper	Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions
Date	24 Feb 2021
Code	https://github.com/whai362/PVT
Authors	Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, Ling Shao
Org.	Nanjing University / The University of Hong Kong / Nanjing University of Science and Technology / IIAI / SenseTime Research
Abstract	Although using convolutional neural networks (CNNs) as backbones achieves great successes in computer vision, this work investigates a simple backbone network useful for many dense prediction tasks without convolutions. Unlike the recently-proposed Transformer model (e.g., ViT) that is specially designed for image classification, we propose Pyramid Vision Transformer~(PVT), which overcomes the difficulties of porting Transformer to various dense prediction tasks. PVT has several merits compared to prior arts. (1) Different from ViT that typically has low-resolution outputs and high computational and memory cost, PVT can be not only trained on dense partitions of the image to achieve high output resolution, which is important for dense predictions but also using a progressive shrinking pyramid to reduce computations of large feature maps. (2) PVT inherits the advantages from both CNN and Transformer, making it a unified backbone in various vision tasks without convolutions by simply replacing CNN backbones. (3) We validate PVT by conducting extensive experiments, showing that it boosts the performance of many downstream tasks, e.g., object detection, semantic, and instance segmentation. For example, with a comparable number of parameters, RetinaNet+PVT achieves 40.4 AP on the COCO dataset, surpassing RetinNet+ResNet50 (36.3 AP) by 4.1 absolute AP. We hope PVT could serve as an alternative and useful backbone for pixel-level predictions and facilitate future researches.

TNT(Transformer-in-Transformer)

Paper	Transformer in Transformer
Date	27 Feb 2021
Code	https://github.com/huawei-noah/noah-research/tree/master/TNT
Authors	Kai Han, An Xiao, Enhua Wu, Jianyuan Guo, Chunjing Xu, Yunhe Wang
Org.	Noah’s Ark Lab, Huawei Technologies / State Key Lab of Computer Science, ISCAS & UCAS
Abstract	Transformer is a type of self-attention-based neural networks originally applied for NLP tasks. Recently, pure transformer-based models are proposed to solve computer vision problems. These visual transformers usually view an image as a sequence of patches while they ignore the intrinsic structure information inside each patch. In this paper, we propose a novel Transformer-iN-Transformer (TNT) model for modeling both patch-level and pixel-level representation. In each TNT block, an outer transformer block is utilized to process patch embeddings, and an inner transformer block extracts local features from pixel embeddings. The pixel-level feature is projected to the space of patch embedding by a linear transformation layer and then added into the patch. By stacking the TNT blocks, we build the TNT model for image recognition. Experiments on ImageNet benchmark and downstream tasks demonstrate the superiority and efficiency of the proposed TNT architecture. For example, our TNT achieves 81.3% top-1 accuracy on ImageNet which is 1.5% higher than that of DeiT with similar computational cost.

Perceiver

Paper	Perceiver: General Perception with Iterative Attention
Date	4 Mar 2021
Code	https://github.com/deepmind/deepmind-research/tree/master/perceiver
Authors	Andrew Jaegle, Felix Gimeno, Andrew Brock, Andrew Zisserman, Oriol Vinyals, Joao Carreira
Org.	DeepMind
Abstract	Biological systems perceive the world by simultaneously processing high-dimensional inputs from modalities as diverse as vision, audition, touch, proprioception, etc. The perception models used in deep learning on the other hand are designed for individual modalities, often relying on domain-specific assumptions such as the local grid structures exploited by virtually all existing vision models. These priors introduce helpful inductive biases, but also lock models to individual modalities. In this paper we introduce the Perceiver - a model that builds upon Transformers and hence makes few architectural assumptions about the relationship between its inputs, but that also scales to hundreds of thousands of inputs, like ConvNets. The model leverages an asymmetric attention mechanism to iteratively distill inputs into a tight latent bottleneck, allowing it to scale to handle very large inputs. We show that this architecture is competitive with or outperforms strong, specialized models on classification tasks across various modalities: images, point clouds, audio, video, and video+audio. The Perceiver obtains performance comparable to ResNet-50 and ViT on ImageNet without 2D convolutions by directly attending to 50,000 pixels. It is also competitive in all modalities in AudioSet.

ConViT(Convolutional Vision Transformer)

Paper	ConViT: Improving Vision Transformers with Soft Convolutional Inductive Biases
Date	19 Mar 2021
Code	https://github.com/facebookresearch/convit
Authors	Stéphane d'Ascoli, Hugo Touvron, Matthew Leavitt, Ari Morcos, Giulio Biroli, Levent Sagun
Org.	Department of Physics, Ecole Normale Superieure / Facebook AI Research
Abstract	Convolutional architectures have proven extremely successful for vision tasks. Their hard inductive biases enable sample-efficient learning, but come at the cost of a potentially lower performance ceiling. Vision Transformers (ViTs) rely on more flexible self-attention layers, and have recently outperformed CNNs for image classification. However, they require costly pre-training on large external datasets or distillation from pre-trained convolutional networks. In this paper, we ask the following question: is it possible to combine the strengths of these two architectures while avoiding their respective limitations? To this end, we introduce gated positional self-attention (GPSA), a form of positional self-attention which can be equipped with a ``soft" convolutional inductive bias. We initialise the GPSA layers to mimic the locality of convolutional layers, then give each attention head the freedom to escape locality by adjusting a gating parameter regulating the attention paid to position versus content information. The resulting convolutional-like ViT architecture, ConViT, outperforms the DeiT on ImageNet, while offering a much improved sample efficiency. We further investigate the role of locality in learning by first quantifying how it is encouraged in vanilla self-attention layers, then analysing how it is escaped in GPSA layers. We conclude by presenting various ablations to better understand the success of the ConViT.

HVT(Hierarchical Visual Transformer)

Paper	Scalable Visual Transformers with Hierarchical Pooling
Date	19 Mar 2021
Code	https://github.com/MonashAI/HVT
Authors	Zizheng Pan, Bohan Zhuang, Jing Liu, Haoyu He, Jianfei Cai
Org.	Dept of Data Science and AI, Faculty of IT, Monash University
Abstract	The recently proposed Visual image Transformers (ViT) with pure attention have achieved promising performance on image recognition tasks, such as image classification. However, the routine of the current ViT model is to maintain a full-length patch sequence during inference, which is redundant and lacks hierarchical representation. To this end, we propose a Hierarchical Visual Transformer (HVT) which progressively pools visual tokens to shrink the sequence length and hence reduces the computational cost, analogous to the feature maps downsampling in Convolutional Neural Networks (CNNs). It brings a great benefit that we can increase the model capacity by scaling dimensions of depth/width/resolution/patch size without introducing extra computational complexity due to the reduced sequence length. Moreover, we empirically find that the average pooled visual tokens contain more discriminative information than the single class token. To demonstrate the improved scalability of our HVT, we conduct extensive experiments on the image classification task. With comparable FLOPs, our HVT outperforms the competitive baselines on ImageNet and CIFAR-100 datasets.

DeepViT

Paper	DeepViT: Towards Deeper Vision Transformer
Date	22 Mar 2021
Code	https://github.com/zhoudaquan/dvit_repo
Authors	Daquan Zhou, Bingyi Kang, Xiaojie Jin, Linjie Yang, Xiaochen Lian, Zihang Jiang, Qibin Hou, Jiashi Feng
Org.	National University of Singapore, ByteDance US AI Lab
Abstract	Vision transformers (ViTs) have been successfully applied in image classification tasks recently. In this paper, we show that, unlike convolution neural networks (CNNs)that can be improved by stacking more convolutional layers, the performance of ViTs saturate fast when scaled to be deeper. More specifically, we empirically observe that such scaling difficulty is caused by the attention collapse issue: as the transformer goes deeper, the attention maps gradually become similar and even much the same after certain layers. In other words, the feature maps tend to be identical in the top layers of deep ViT models. This fact demonstrates that in deeper layers of ViTs, the self-attention mechanism fails to learn effective concepts for representation learning and hinders the model from getting expected performance gain. Based on above observation, we propose a simple yet effective method, named Re-attention, to re-generate the attention maps to increase their diversity at different layers with negligible computation and memory cost. The pro-posed method makes it feasible to train deeper ViT models with consistent performance improvements via minor modification to existing ViT models. Notably, when training a deep ViT model with 32 transformer blocks, the Top-1 classification accuracy can be improved by 1.6% on ImageNet.

HaloNet

Paper	Scaling Local Self-Attention for Parameter Efficient Visual Backbones
Date	23 Mar 2021
Code
Authors	Ashish Vaswani, Prajit Ramachandran, Aravind Srinivas, Niki Parmar, Blake Hechtman, Jonathon Shlens
Org.	Google Research, UC Berkeley
Abstract	Self-attention has the promise of improving computer vision systems due to parameter-independent scaling of receptive fields and content-dependent interactions, in contrast to parameter-dependent scaling and content-independent interactions of convolutions. Self-attention models have recently been shown to have encouraging improvements on accuracy-parameter trade-offs compared to baseline convolutional models such as ResNet-50. In this work, we aim to develop self-attention models that can outperform not just the canonical baseline models, but even the high-performing convolutional models. We propose two extensions to self-attention that, in conjunction with a more efficient implementation of self-attention, improve the speed, memory usage, and accuracy of these models. We leverage these improvements to develop a new self-attention model family, HaloNets, which reach state-of-the-art accuracies on the parameter-limited setting of the ImageNet classification benchmark. In preliminary transfer learning experiments, we find that HaloNet models outperform much larger models and have better inference performance. On harder tasks such as object detection and instance segmentation, our simple local self-attention and convolutional hybrids show improvements over very strong baselines. These results mark another step in demonstrating the efficacy of self-attention models on settings traditionally dominated by convolutional models.

Swin(Shifted Windows Transformer)

Paper	Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
Date	25 Mar 2021
Code	https://github.com/microsoft/Swin-Transformer
Authors	Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, Baining Guo
Org.	Microsoft Research Asia
Abstract	This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (86.4 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones.

CvT(Convolutional vision Transformer)

Paper	CvT: Introducing Convolutions to Vision Transformers
Date	29 Mar 2021
Code	https://github.com/leoxiaobin/CvT
Authors	Haiping Wu, Bin Xiao, Noel Codella, Mengchen Liu, Xiyang Dai, Lu Yuan, Lei Zhang
Org.	McGill University, Microsoft Cloud + AI
Abstract	We present in this paper a new architecture, named Convolutional vision Transformer (CvT), that improves Vision Transformer (ViT) in performance and efficiency by introducing convolutions into ViT to yield the best of both designs. This is accomplished through two primary modifications: a hierarchy of Transformers containing a new convolutional token embedding, and a convolutional Transformer block leveraging a convolutional projection. These changes introduce desirable properties of convolutional neural networks (CNNs) to the ViT architecture (\ie shift, scale, and distortion invariance) while maintaining the merits of Transformers (\ie dynamic attention, global context, and better generalization). We validate CvT by conducting extensive experiments, showing that this approach achieves state-of-the-art performance over other Vision Transformers and ResNets on ImageNet-1k, with fewer parameters and lower FLOPs. In addition, performance gains are maintained when pretrained on larger datasets (\eg ImageNet-22k) and fine-tuned to downstream tasks. Pre-trained on ImageNet-22k, our CvT-W24 obtains a top-1 accuracy of 87.7% on the ImageNet-1k val set. Finally, our results show that the positional encoding, a crucial component in existing Vision Transformers, can be safely removed in our model, simplifying the design for higher resolution vision tasks.

ViL(Vision Longformer)

Paper	Multi-Scale Vision Longformer: A New Vision Transformer for High-Resolution Image Encoding
Date	29 Mar 2021
Code	https://github.com/microsoft/vision-longformer
Authors	Pengchuan Zhang, Xiyang Dai, Jianwei Yang, Bin Xiao, Lu Yuan, Lei Zhang, Jianfeng Gao
Org.	Microsoft Corporation
Abstract	This paper presents a new Vision Transformer (ViT) architecture Multi-Scale Vision Longformer, which significantly enhances the ViT of \cite{dosovitskiy2020image} for encoding high-resolution images using two techniques. The first is the multi-scale model structure, which provides image encodings at multiple scales with manageable computational cost. The second is the attention mechanism of vision Longformer, which is a variant of Longformer \cite{beltagy2020longformer}, originally developed for natural language processing, and achieves a linear complexity w.r.t. the number of input tokens. A comprehensive empirical study shows that the new ViT significantly outperforms several strong baselines, including the existing ViT models and their ResNet counterparts, and the Pyramid Vision Transformer from a concurrent work \cite{wang2021pyramid}, on a range of vision tasks, including image classification, object detection, and segmentation.

PiT(Pooling-based Vision Transformer)

Paper	Rethinking Spatial Dimensions of Vision Transformers
Date	30 Mar 2021
Code	https://github.com/naver-ai/pit
Authors	Byeongho Heo, Sangdoo Yun, Dongyoon Han, Sanghyuk Chun, Junsuk Choe, Seong Joon Oh
Org.	NAVER AI Lab
Abstract	Vision Transformer (ViT) extends the application range of transformers from language processing to computer vision tasks as being an alternative architecture against the existing convolutional neural networks (CNN). Since the transformer-based architecture has been innovative for computer vision modeling, the design convention towards an effective architecture has been less studied yet. From the successful design principles of CNN, we investigate the role of the spatial dimension conversion and its effectiveness on the transformer-based architecture. We particularly attend the dimension reduction principle of CNNs; as the depth increases, a conventional CNN increases channel dimension and decreases spatial dimensions. We empirically show that such a spatial dimension reduction is beneficial to a transformer architecture as well, and propose a novel Pooling-based Vision Transformer (PiT) upon the original ViT model. We show that PiT achieves the improved model capability and generalization performance against ViT. Throughout the extensive experiments, we further show PiT outperforms the baseline on several tasks such as image classification, object detection and robustness evaluation.

Cait(Class-Attention in Image Transformers)

Paper	Going deeper with Image Transformers
Date	31 Mar 2021
Code	https://github.com/facebookresearch/deit
Authors	Hugo Touvron, Matthieu Cord, Alexandre Sablayrolles, Gabriel Synnaeve, Hervé Jégou
Org.	Facebook AI, Sorbonne University
Abstract	Transformers have been recently adapted for large scale image classification, achieving high scores shaking up the long supremacy of convolutional neural networks. However the optimization of image transformers has been little studied so far. In this work, we build and optimize deeper transformer networks for image classification. In particular, we investigate the interplay of architecture and optimization of such dedicated transformers. We make two transformers architecture changes that significantly improve the accuracy of deep transformers. This leads us to produce models whose performance does not saturate early with more depth, for instance we obtain 86.5% top-1 accuracy on Imagenet when training with no external data, we thus attain the current SOTA with less FLOPs and parameters. Moreover, our best model establishes the new state of the art on Imagenet with Reassessed labels and Imagenet-V2 / match frequency, in the setting with no additional training data. We share our code and models.

2021 Q2

Visformer(Vision-friendly Transformer)

Paper	Visformer: The Vision-friendly Transformer
Date	6 Apr 2021
Code	https://github.com/danczs/Visformer
Authors	Zhengsu Chen, Lingxi Xie, Jianwei Niu, Xuefeng Liu, Longhui Wei, Qi Tian
Org.	Beihang University, Johns Hopkins University, Peking University, Xidian University
Abstract	The past year has witnessed the rapid development of applying the Transformer module to vision problems. While some researchers have demonstrated that Transformer-based models enjoy a favorable ability of fitting data, there are still growing number of evidences showing that these models suffer over-fitting especially when the training data is limited. This paper offers an empirical study by performing step-by-step operations to gradually transit a Transformer-based model to a convolution-based model. The results we obtain during the transition process deliver useful messages for improving visual recognition. Based on these observations, we propose a new architecture named Visformer, which is abbreviated from the `Vision-friendly Transformer'. With the same computational complexity, Visformer outperforms both the Transformer-based and convolution-based models in terms of ImageNet classification accuracy, and the advantage becomes more significant when the model complexity is lower or the training set is smaller.

SiT(Self-supervised vision Transformers)

Paper	SiT: Self-supervised vIsion Transformer
Date	8 Apr 2021
Code	https://github.com/Sara-Ahmed/SiT
Authors	Sara Atito, Muhammad Awais, Josef Kittler
Org.	Centre for Vision, Speech and Signal Processing (CVSSP), University of Surrey
Abstract	Self-supervised learning methods are gaining increasing traction in computer vision due to their recent success in reducing the gap with supervised learning. In natural language processing (NLP) self-supervised learning and transformers are already the methods of choice. The recent literature suggests that the transformers are becoming increasingly popular also in computer vision. So far, the vision transformers have been shown to work well when pretrained either using a large scale supervised data or with some kind of co-supervision, e.g. in terms of teacher network. These supervised pretrained vision transformers achieve very good results in downstream tasks with minimal changes. In this work we investigate the merits of self-supervised learning for pretraining image/vision transformers and then using them for downstream classification tasks. We propose Self-supervised vIsion Transformers (SiT) and discuss several self-supervised training mechanisms to obtain a pretext model. The architectural flexibility of SiT allows us to use it as an autoencoder and work with multiple self-supervised tasks seamlessly. We show that a pretrained SiT can be finetuned for a downstream classification task on small scale datasets, consisting of a few thousand images rather than several millions. The proposed approach is evaluated on standard datasets using common protocols. The results demonstrate the strength of the transformers and their suitability for self-supervised learning. We outperformed existing self-supervised learning methods by large margin. We also observed that SiT is good for few shot learning and also showed that it is learning useful representation by simply training a linear classifier on top of the learned features from SiT.

LocalVit(Locality Vision Transformers)

Paper	LocalViT: Bringing Locality to Vision Transformers
Date	12 Apr 2021
Code	https://github.com/ofsoundof/LocalViT
Authors	Yawei Li, Kai Zhang, Jiezhang Cao, Radu Timofte, Luc Van Gool
Org.	Computer Vision Lab, ETH Zurich, Switzerland / KU Leuven, Belgium
Abstract	We study how to introduce locality mechanisms into vision transformers. The transformer network originates from machine translation and is particularly good at modelling long-range dependencies within a long sequence. Although the global interaction between the token embeddings could be well modelled by the self-attention mechanism of transformers, what is lacking a locality mechanism for information exchange within a local region. Yet, locality is essential for images since it pertains to structures like lines, edges, shapes, and even objects.We add locality to vision transformers by introducing depth-wise convolution into the feed-forward network. This seemingly simple solution is inspired by the comparison between feed-forward networks and inverted residual blocks. The importance of locality mechanisms is validated in two ways: 1) A wide range of design choices (activation function, layer placement, expansion ratio) are available for incorporating locality mechanisms and all proper choices can lead to a performance gain over the baseline, and 2) The same locality mechanism is successfully applied to 4 vision transformers, which shows the generalization of the locality concept. In particular, for ImageNet2012 classification, the locality-enhanced transformers outperform the baselines DeiT-T and PVT-T by 2.6% and 3.1% with a negligible increase in the number of parameters and computational effort.

CCT(Compact Convolutional Transformer)

Paper	Escaping the Big Data Paradigm with Compact Transformers
Date	12 Apr 2021
Code	https://github.com/SHI-Labs/Compact-Transformers
Authors	Ali Hassani, Steven Walton, Nikhil Shah, Abulikemu Abuduweili, Jiachen Li, Humphrey Shi
Org.	University of Oregon / University of Illinois at Urbana-Champaign / Picsart AI Research (PAIR)
Abstract	With the rise of Transformers as the standard for language processing, and their advancements in computer vision, along with their unprecedented size and amounts of training data, many have come to believe that they are not suitable for small sets of data. This trend leads to great concerns, including but not limited to: limited availability of data in certain scientific domains and the exclusion of those with limited resource from research in the field. In this paper, we dispel the myth that transformers are "data hungry" and therefore can only be applied to large sets of data. We show for the first time that with the right size and tokenization, transformers can perform head-to-head with state-of-the-art CNNs on small datasets. Our model eliminates the requirement for class token and positional embeddings through a novel sequence pooling strategy and the use of convolutions. We show that compared to CNNs, our compact transformers have fewer parameters and MACs, while obtaining similar accuracies. Our method is flexible in terms of model size, and can have as little as 0.28M parameters and achieve reasonable results. It can reach an accuracy of 95.29 % when training from scratch on CIFAR-10, which is comparable with modern CNN based approaches, and a significant improvement over previous Transformer based models. Our simple and compact design democratizes transformers by making them accessible to those equipped with basic computing resources and/or dealing with important small datasets. Our method works on larger datasets, such as ImageNet (80.28% accuracy with 29% parameters of ViT), and NLP tasks as well.

CoaT(Co-scale conv-attentional image Transformers)

Paper	Co-Scale Conv-Attentional Image Transformers
Date	13 Apr 2021
Code	https://github.com/mlpc-ucsd/CoaT
Authors	Weijian Xu, Yifan Xu, Tyler Chang, Zhuowen Tu
Org.	University of California San Diego
Abstract	In this paper, we present Co-scale conv-attentional image Transformers (CoaT), a Transformer-based image classifier equipped with co-scale and conv-attentional mechanisms. First, the co-scale mechanism maintains the integrity of Transformers' encoder branches at individual scales, while allowing representations learned at different scales to effectively communicate with each other; we design a series of serial and parallel blocks to realize the co-scale attention mechanism. Second, we devise a conv-attentional mechanism by realizing a relative position embedding formulation in the factorized attention module with an efficient convolution-like implementation. CoaT empowers image Transformers with enriched multi-scale and contextual modeling capabilities. On ImageNet, relatively small CoaT models attain superior classification results compared with the similar-sized convolutional neural networks and image/vision Transformers. The effectiveness of CoaT's backbone is also illustrated on object detection and instance segmentation, demonstrating its applicability to the downstream computer vision tasks.

MViTv1(Multiscale Vision Transformers)

Paper	Multiscale Vision Transformers
Date	22 Apr 2021
Code	https://github.com/facebookresearch/SlowFast
Authors	Haoqi Fan, Bo Xiong, Karttikeya Mangalam, Yanghao Li, Zhicheng Yan, Jitendra Malik, Christoph Feichtenhofer
Org.	Facebook AI Research , UC Berkeley
Abstract	We present Multiscale Vision Transformers (MViT) for video and image recognition, by connecting the seminal idea of multiscale feature hierarchies with transformer models. Multiscale Transformers have several channel-resolution scale stages. Starting from the input resolution and a small channel dimension, the stages hierarchically expand the channel capacity while reducing the spatial resolution. This creates a multiscale pyramid of features with early layers operating at high spatial resolution to model simple low-level visual information, and deeper layers at spatially coarse, but complex, high-dimensional features. We evaluate this fundamental architectural prior for modeling the dense nature of visual signals for a variety of video recognition tasks where it outperforms concurrent vision transformers that rely on large scale external pre-training and are 5-10x more costly in computation and parameters. We further remove the temporal dimension and apply our model for image classification where it outperforms prior work on vision transformers.

ConTNet(ConvolutionTransformer Network)

Paper	ConTNet: Why not use convolution and transformer at the same time?
Date	27 Apr 2021
Code	https://github.com/yan-hao-tian/ConTNet
Authors	Haotian Yan, Zhe Li, Weijian Li, Changhu Wang, Ming Wu, Chuang Zhang
Org.	School of AI, Beijing University of Posts and Telecommunications. / ByteDance AI Lab, Beijing.
Abstract	Although convolutional networks (ConvNets) have enjoyed great success in computer vision (CV), it suffers from capturing global information crucial to dense prediction tasks such as object detection and segmentation. In this work, we innovatively propose ConTNet (ConvolutionTransformer Network), combining transformer with ConvNet architectures to provide large receptive fields. Unlike the recently-proposed transformer-based models (e.g., ViT, DeiT) that are sensitive to hyper-parameters and extremely dependent on a pile of data augmentations when trained from scratch on a midsize dataset (e.g., ImageNet1k), ConTNet can be optimized like normal ConvNets (e.g., ResNet) and preserve an outstanding robustness. It is also worth pointing that, given identical strong data augmentations, the performance improvement of ConTNet is more remarkable than that of ResNet. We present its superiority and effectiveness on image classification and downstream tasks. For example, our ConTNet achieves 81.8% top-1 accuracy on ImageNet which is the same as DeiT-B with less than 40% computational complexity. ConTNet-M also outperforms ResNet50 as the backbone of both Faster-RCNN (by 2.6%) and Mask-RCNN (by 3.2%) on COCO2017 dataset. We hope that ConTNet could serve as a useful backbone for CV tasks and bring new ideas for model design

Twins

Paper	Twins: Revisiting the Design of Spatial Attention in Vision Transformers
Date	28 Apr 2021
Code	https://github.com/Meituan-AutoML/Twins
Authors	Xiangxiang Chu, Zhi Tian, Yuqing Wang, Bo Zhang, Haibing Ren, Xiaolin Wei, Huaxia Xia, Chunhua Shen
Org.	Meituan Inc., The University of Adelaide
Abstract	Very recently, a variety of vision transformer architectures for dense prediction tasks have been proposed and they show that the design of spatial attention is critical to their success in these tasks. In this work, we revisit the design of the spatial attention and demonstrate that a carefully-devised yet simple spatial attention mechanism performs favourably against the state-of-the-art schemes. As a result, we propose two vision transformer architectures, namely, Twins-PCPVT and Twins-SVT. Our proposed architectures are highly-efficient and easy to implement, only involving matrix multiplications that are highly optimized in modern deep learning frameworks. More importantly, the proposed architectures achieve excellent performance on a wide range of visual tasks, including image level classification as well as dense detection and segmentation. The simplicity and strong performance suggest that our proposed architectures may serve as stronger backbones for many vision tasks.

Conformer

Paper	Conformer: Local Features Coupling Global Representations for Visual Recognition
Date	9 May 2021
Code	https://github.com/pengzhiliang/Conformer
Authors	Zhiliang Peng, Wei Huang, Shanzhi Gu, Lingxi Xie, Yaowei Wang, Jianbin Jiao, Qixiang Ye
Org.	University of Chinese Academy of Sciences, Huawei Inc. ,Peng Cheng Laboratory
Abstract	Within Convolutional Neural Network (CNN), the convolution operations are good at extracting local features but experience difficulty to capture global representations. Within visual transformer, the cascaded self-attention modules can capture long-distance feature dependencies but unfortunately deteriorate local feature details. In this paper, we propose a hybrid network structure, termed Conformer, to take advantage of convolutional operations and self-attention mechanisms for enhanced representation learning. Conformer roots in the Feature Coupling Unit (FCU), which fuses local features and global representations under different resolutions in an interactive fashion. Conformer adopts a concurrent structure so that local features and global representations are retained to the maximum extent. Experiments show that Conformer, under the comparable parameter complexity, outperforms the visual transformer (DeiT-B) by 2.3% on ImageNet. On MSCOCO, it outperforms ResNet-101 by 3.7% and 3.6% mAPs for object detection and instance segmentation, respectively, demonstrating the great potential to be a general backbone network.

NesT(Nested Transformers)

Paper	Aggregating Nested Transformers
Date	26 May 2021
Code	https://github.com/google-research/nested-transformer
Authors	Zizhao Zhang, Han Zhang, Long Zhao, Ting Chen, Tomas Pfister
Org.	Google Cloud AI, Google Research, Rutgers University
Abstract	Although hierarchical structures are popular in recent vision transformers, they require sophisticated designs and massive datasets to work well. In this work, we explore the idea of nesting basic local transformers on non-overlapping image blocks and aggregating them in a hierarchical manner. We find that the block aggregation function plays a critical role in enabling cross-block non-local information communication. This observation leads us to design a simplified architecture with minor code changes upon the original vision transformer and obtains improved performance compared to existing methods. Our empirical results show that the proposed method NesT converges faster and requires much less training data to achieve good generalization. For example, a NesT with 68M parameters trained on ImageNet for 100/300 epochs achieves $82.3%/83.8%$ accuracy evaluated on $224\times 224$ image size, outperforming previous methods with up to $57%$ parameter reduction. Training a NesT with 6M parameters from scratch on CIFAR10 achieves $96%$ accuracy using a single GPU, setting a new state of the art for vision transformers. Beyond image classification, we extend the key idea to image generation and show NesT leads to a strong decoder that is 8$\times$ faster than previous transformer based generators. Furthermore, we also propose a novel method for visually interpreting the learned model.

ResT(ResNet Transformer)

Paper	ResT: An Efficient Transformer for Visual Recognition
Date	28 May 2021
Code	https://github.com/wofmanaf/ResT
Authors	Qinglong Zhang, Yubin Yang
Org.	State Key Laboratory for Novel Software Technology / Nanjing University
Abstract	This paper presents an efficient multi-scale vision Transformer, called ResT, that capably served as a general-purpose backbone for image recognition. Unlike existing Transformer methods, which employ standard Transformer blocks to tackle raw images with a fixed resolution, our ResT have several advantages: (1) A memory-efficient multi-head self-attention is built, which compresses the memory by a simple depth-wise convolution, and projects the interaction across the attention-heads dimension while keeping the diversity ability of multi-heads; (2) Position encoding is constructed as spatial attention, which is more flexible and can tackle with input images of arbitrary size without interpolation or fine-tune; (3) Instead of the straightforward tokenization at the beginning of each stage, we design the patch embedding as a stack of overlapping convolution operation with stride on the 2D-reshaped token map. We comprehensively validate ResT on image classification and downstream tasks. Experimental results show that the proposed ResT can outperform the recently state-of-the-art backbones by a large margin, demonstrating the potential of ResT as strong backbones.

LIT(Less attention vIsion Transformer)

Paper	Less is More: Pay Less Attention in Vision Transformers
Date	29 May 2021
Code	https://github.com/MonashAI/LIT
Authors	Zizheng Pan, Bohan Zhuang, Haoyu He, Jing Liu, Jianfei Cai
Org.	Dept of Data Science and AI, Monash University
Abstract	Transformers have become one of the dominant architectures in deep learning, particularly as a powerful alternative to convolutional neural networks (CNNs) in computer vision. However, Transformer training and inference in previous works can be prohibitively expensive due to the quadratic complexity of self-attention over a long sequence of representations, especially for high-resolution dense prediction tasks. To this end, we present a novel Less attention vIsion Transformer (LIT), building upon the fact that convolutions, fully-connected (FC) layers, and self-attentions have almost equivalent mathematical expressions for processing image patch sequences. Specifically, we propose a hierarchical Transformer where we use pure multi-layer perceptrons (MLPs) to encode rich local patterns in the early stages while applying self-attention modules to capture longer dependencies in deeper layers. Moreover, we further propose a learned deformable token merging module to adaptively fuse informative patches in a non-uniform manner. The proposed LIT achieves promising performance on image recognition tasks, including image classification, object detection and instance segmentation, serving as a strong backbone for many vision tasks.

DVT(Dynamic Vision Transformers)

Paper	Not All Images are Worth 16x16 Words: Dynamic Vision Transformers with Adaptive Sequence Length
Date	31 May 2021
Code	https://github.com/blackfeather-wang/Dynamic-Vision-Transformer
Authors	Yulin Wang, Rui Huang, Shiji Song, Zeyi Huang, Gao Huang
Org.	Tsinghua University, Huawei Technologies Ltd.
Abstract	Vision Transformers (ViT) have achieved remarkable success in large-scale image recognition. They split every 2D image into a fixed number of patches, each of which is treated as a token. Generally, representing an image with more tokens would lead to higher prediction accuracy, while it also results in drastically increased computational cost. To achieve a decent trade-off between accuracy and speed, the number of tokens is empirically set to 16x16. In this paper, we argue that every image has its own characteristics, and ideally the token number should be conditioned on each individual input. In fact, we have observed that there exist a considerable number of "easy" images which can be accurately predicted with a mere number of 4x4 tokens, while only a small fraction of "hard" ones need a finer representation. Inspired by this phenomenon, we propose a Dynamic Transformer to automatically configure a proper number of tokens for each input image. This is achieved by cascading multiple Transformers with increasing numbers of tokens, which are sequentially activated in an adaptive fashion at test time, i.e., the inference is terminated once a sufficiently confident prediction is produced. We further design efficient feature reuse and relationship reuse mechanisms across different components of the Dynamic Transformer to reduce redundant computations. Extensive empirical results on ImageNet, CIFAR-10, and CIFAR-100 demonstrate that our method significantly outperforms the competitive baselines in terms of both theoretical computational efficiency and practical inference speed.

MSG-Transformer(Messanger Transformer)

Paper	MSG-Transformer: Exchanging Local Spatial Information by Manipulating Messenger Tokens
Date	31 May 2021
Code	https://github.com/hustvl/MSG-Transformer
Authors	Jiemin Fang, Lingxi Xie, Xinggang Wang, Xiaopeng Zhang, Wenyu Liu, Qi Tian
Org.	Huazhong University of Science & Technology, Huawei Inc.
Abstract	Transformers have offered a new methodology of designing neural networks for visual recognition. Compared to convolutional networks, Transformers enjoy the ability of referring to global features at each stage, yet the attention module brings higher computational overhead that obstructs the application of Transformers to process high-resolution visual data. This paper aims to alleviate the conflict between efficiency and flexibility, for which we propose a specialized token for each region that serves as a messenger (MSG). Hence, by manipulating these MSG tokens, one can flexibly exchange visual information across regions and the computational complexity is reduced. We then integrate the MSG token into a multi-scale architecture named MSG-Transformer. In standard image classification and object detection, MSG-Transformer achieves competitive performance and the inference on both GPU and CPU is accelerated.

TransCNN

Paper	Transformer in Convolutional Neural Networks
Date	6 Jun 2021
Code	https://github.com/yun-liu/TransCNN
Authors	Yun Liu, Guolei Sun, Yu Qiu, Le Zhang, Ajad Chhatkuli, Luc Van Gool
Org.	CVL, ETH Zurich / Nankai University / UESTC
Abstract	We tackle the low-efficiency flaw of vision transformer caused by the high computational/space complexity in Multi-Head Self-Attention (MHSA). To this end, we propose the Hierarchical MHSA (H-MHSA), whose representation is computed in a hierarchical manner. Specifically, our H-MHSA first learns feature relationships within small grids by viewing image patches as tokens. Then, small grids are merged into larger ones, within which feature relationship is learned by viewing each small grid at the preceding step as a token. This process is iterated to gradually reduce the number of tokens. The H-MHSA module is readily pluggable into any CNN architectures and amenable to training via backpropagation. We call this new backbone TransCNN, and it essentially inherits the advantages of both transformer and CNN. Experiments demonstrate that TransCNN achieves state-of-the-art accuracy for image recognition. Code and pretrained models are available at this https URL. This technical report will keep updating by adding more experiments.

Refiner(Refined Vision Transformer)

Paper	Refiner: Refining Self-attention for Vision Transformers
Date	7 Jun 2021
Code	https://github.com/zhoudaquan/Refiner_ViT
Authors	Daquan Zhou, Yujun Shi, Bingyi Kang, Weihao Yu, Zihang Jiang, Yuan Li, Xiaojie Jin, Qibin Hou, Jiashi Feng
Org.	National University of Singapore
Abstract	Vision Transformers (ViTs) have shown competitive accuracy in image classification tasks compared with CNNs. Yet, they generally require much more data for model pre-training. Most of recent works thus are dedicated to designing more complex architectures or training methods to address the data-efficiency issue of ViTs. However, few of them explore improving the self-attention mechanism, a key factor distinguishing ViTs from CNNs. Different from existing works, we introduce a conceptually simple scheme, called refiner, to directly refine the self-attention maps of ViTs. Specifically, refiner explores attention expansion that projects the multi-head attention maps to a higher-dimensional space to promote their diversity. Further, refiner applies convolutions to augment local patterns of the attention maps, which we show is equivalent to a distributed local attention features are aggregated locally with learnable kernels and then globally aggregated with self-attention. Extensive experiments demonstrate that refiner works surprisingly well. Significantly, it enables ViTs to achieve 86% top-1 classification accuracy on ImageNet with only 81M parameters.

Shuffle Transformer

Paper	Shuffle Transformer: Rethinking Spatial Shuffle for Vision Transformer
Date	7 Jun 2021
Code	https://github.com/speedinghzl/ShuffleTransformer
Authors	Zilong Huang, Youcheng Ben, Guozhong Luo, Pei Cheng, Gang Yu, Bin Fu
Org.	Tencent GY-Lab
Abstract	Very recently, Window-based Transformers, which computed self-attention within non-overlapping local windows, demonstrated promising results on image classification, semantic segmentation, and object detection. However, less study has been devoted to the cross-window connection which is the key element to improve the representation ability. In this work, we revisit the spatial shuffle as an efficient way to build connections among windows. As a result, we propose a new vision transformer, named Shuffle Transformer, which is highly efficient and easy to implement by modifying two lines of code. Furthermore, the depth-wise convolution is introduced to complement the spatial shuffle for enhancing neighbor-window connections. The proposed architectures achieve excellent performance on a wide range of visual tasks including image-level classification, object detection, and semantic segmentation. Code will be released for reproduction.

ViTAE(Vision Transformer Advanced by Exploring Intrinsic Inductive Bias)

Paper	ViTAE: Vision Transformer Advanced by Exploring Intrinsic Inductive Bias
Date	7 Jun 2021
Code	https://github.com/Annbless/ViTAE
Authors	Yufei Xu, Qiming Zhang, Jing Zhang, Dacheng Tao
Org.	The University of Sydney, JD Explore Academy
Abstract	Transformers have shown great potential in various computer vision tasks owing to their strong capability in modeling long-range dependency using the self-attention mechanism. Nevertheless, vision transformers treat an image as 1D sequence of visual tokens, lacking an intrinsic inductive bias (IB) in modeling local visual structures and dealing with scale variance. Alternatively, they require large-scale training data and longer training schedules to learn the IB implicitly. In this paper, we propose a novel Vision Transformer Advanced by Exploring intrinsic IB from convolutions, \ie, ViTAE. Technically, ViTAE has several spatial pyramid reduction modules to downsample and embed the input image into tokens with rich multi-scale context by using multiple convolutions with different dilation rates. In this way, it acquires an intrinsic scale invariance IB and is able to learn robust feature representation for objects at various scales. Moreover, in each transformer layer, ViTAE has a convolution block in parallel to the multi-head self-attention module, whose features are fused and fed into the feed-forward network. Consequently, it has the intrinsic locality IB and is able to learn local features and global dependencies collaboratively. Experiments on ImageNet as well as downstream tasks prove the superiority of ViTAE over the baseline transformer and concurrent works.

CoAtNet(Convolution and self-Attention Net)

Paper	CoAtNet: Marrying Convolution and Attention for All Data Sizes
Date	9 Jun 2021
Code
Authors	Zihang Dai, Hanxiao Liu, Quoc V. Le, Mingxing Tan
Org.	Google Research, Brain Team
Abstract	Transformers have attracted increasing interests in computer vision, but they still fall behind state-of-the-art convolutional networks. In this work, we show that while Transformers tend to have larger model capacity, their generalization can be worse than convolutional networks due to the lack of the right inductive bias. To effectively combine the strengths from both architectures, we present CoAtNets(pronounced "coat" nets), a family of hybrid models built from two key insights:(1) depthwise Convolution and self-Attention can be naturally unified via simple relative attention; (2) vertically stacking convolution layers and attention layers in a principled way is surprisingly effective in improving generalization, capacity and efficiency. Experiments show that our CoAtNets achieve state-of-the-art performance under different resource constraints across various datasets. For example, CoAtNet achieves 86.0% ImageNet top-1 accuracy without extra data, and 89.77% with extra JFT data, outperforming prior arts of both convolutional networks and Transformers. Notably, when pre-trained with 13M images fromImageNet-21K, our CoAtNet achieves 88.56% top-1 accuracy, matching ViT-huge pre-trained with 300M images from JFT while using 23x less data.

CAT(Cross Attention Vision Transformer)

Paper	CAT: Cross Attention in Vision Transformer
Date	10 Jun 2021
Code	https://github.com/linhezheng19/CAT
Authors	Hezheng Lin, Xing Cheng, Xiangyu Wu, Fan Yang, Dong Shen, Zhongyuan Wang, Qing Song, Wei Yuan
Org.	MMU, KuaiShou Inc. , Beijing University of Posts and Telecommunications
Abstract	Since Transformer has found widespread use in NLP, the potential of Transformer in CV has been realized and has inspired many new approaches. However, the computation required for replacing word tokens with image patches for Transformer after the tokenization of the image is vast(e.g., ViT), which bottlenecks model training and inference. In this paper, we propose a new attention mechanism in Transformer termed Cross Attention, which alternates attention inner the image patch instead of the whole image to capture local information and apply attention between image patches which are divided from single-channel feature maps capture global information. Both operations have less computation than standard self-attention in Transformer. By alternately applying attention inner patch and between patches, we implement cross attention to maintain the performance with lower computational cost and build a hierarchical network called Cross Attention Transformer(CAT) for other vision tasks. Our base model achieves state-of-the-arts on ImageNet-1K, and improves the performance of other methods on COCO and ADE20K, illustrating that our network has the potential to serve as general backbones.

V-MoE(Vision Mixture of Experts)

Paper	Scaling Vision with Sparse Mixture of Experts
Date	10 Jun 2021
Code	https://github.com/google-research/vmoe
Authors	Carlos Riquelme, Joan Puigcerver, Basil Mustafa, Maxim Neumann, Rodolphe Jenatton, André Susano Pinto, Daniel Keysers, Neil Houlsby
Org.	Google Brain
Abstract	Sparsely-gated Mixture of Experts networks (MoEs) have demonstrated excellent scalability in Natural Language Processing. In Computer Vision, however, almost all performant networks are "dense", that is, every input is processed by every parameter. We present a Vision MoE (V-MoE), a sparse version of the Vision Transformer, that is scalable and competitive with the largest dense networks. When applied to image recognition, V-MoE matches the performance of state-of-the-art networks, while requiring as little as half of the compute at inference time. Further, we propose an extension to the routing algorithm that can prioritize subsets of each input across the entire batch, leading to adaptive per-image compute. This allows V-MoE to trade-off performance and compute smoothly at test-time. Finally, we demonstrate the potential of V-MoE to scale vision models, and train a 15B parameter model that attains 90.35% on ImageNet.

BEiT(Bidirectional Encoder representation Image Transformers)

Paper	BEiT: BERT Pre-Training of Image Transformers
Date	15 Jun 2021
Code	https://aka.ms/beit
Authors	Hangbo Bao, Li Dong, Furu Wei
Org.	Microsoft Research
Abstract	We introduce a self-supervised vision representation model BEiT, which stands for Bidirectional Encoder representation from Image Transformers. Following BERT developed in the natural language processing area, we propose a masked image modeling task to pretrain vision Transformers. Specifically, each image has two views in our pre-training, i.e, image patches (such as 16x16 pixels), and visual tokens (i.e., discrete tokens). We first "tokenize" the original image into visual tokens. Then we randomly mask some image patches and fed them into the backbone Transformer. The pre-training objective is to recover the original visual tokens based on the corrupted image patches. After pre-training BEiT, we directly fine-tune the model parameters on downstream tasks by appending task layers upon the pretrained encoder. Experimental results on image classification and semantic segmentation show that our model achieves competitive results with previous pre-training methods. For example, base-size BEiT achieves 83.2% top-1 accuracy on ImageNet-1K, significantly outperforming from-scratch DeiT training (81.8%) with the same setup. Moreover, large-size BEiT obtains 86.3% only using ImageNet-1K, even outperforming ViT-L with supervised pre-training on ImageNet-22K (85.2%).

XCiT(Cross-Covariance Image Transformers)

Paper	XCiT: Cross-Covariance Image Transformers
Date	17 Jun 2021
Code	https://github.com/facebookresearch/xcit
Authors	Alaaeldin El-Nouby, Hugo Touvron, Mathilde Caron, Piotr Bojanowski, Matthijs Douze, Armand Joulin, Ivan Laptev, Natalia Neverova, Gabriel Synnaeve, Jakob Verbeek, Hervé Jegou
Org.	Facebook AI , Inria , Sorbonne University
Abstract	Following their success in natural language processing, transformers have recently shown much promise for computer vision. The self-attention operation underlying transformers yields global interactions between all tokens ,i.e. words or image patches, and enables flexible modelling of image data beyond the local interactions of convolutions. This flexibility, however, comes with a quadratic complexity in time and memory, hindering application to long sequences and high-resolution images. We propose a "transposed" version of self-attention that operates across feature channels rather than tokens, where the interactions are based on the cross-covariance matrix between keys and queries. The resulting cross-covariance attention (XCA) has linear complexity in the number of tokens, and allows efficient processing of high-resolution images. Our cross-covariance image transformer (XCiT) is built upon XCA. It combines the accuracy of conventional transformers with the scalability of convolutional architectures. We validate the effectiveness and generality of XCiT by reporting excellent results on multiple vision benchmarks, including image classification and self-supervised feature learning on ImageNet-1k, object detection and instance segmentation on COCO, and semantic segmentation on ADE20k.

EsVit(Efficient Self-supervised Vision Transformers)

Paper	Efficient Self-supervised Vision Transformers for Representation Learning
Date	17 Jun 2021
Code	https://github.com/microsoft/esvit
Authors	Chunyuan Li, Jianwei Yang, Pengchuan Zhang, Mei Gao, Bin Xiao, Xiyang Dai, Lu Yuan, Jianfeng Gao
Org.	Microsoft Research at Redmond, Microsoft Cloud + AI
Abstract	This paper investigates two techniques for developing efficient self-supervised vision transformers (EsViT) for visual representation learning. First, we show through a comprehensive empirical study that multi-stage architectures with sparse self-attentions can significantly reduce modeling complexity but with a cost of losing the ability to capture fine-grained correspondences between image regions. Second, we propose a new pre-training task of region matching which allows the model to capture fine-grained region dependencies and as a result significantly improves the quality of the learned vision representations. Our results show that combining the two techniques, EsViT achieves 81.3% top-1 on the ImageNet linear probe evaluation, outperforming prior arts with around an order magnitude of higher throughput. When transferring to downstream linear classification tasks, EsViT outperforms its supervised counterpart on 17 out of 18 datasets.

P2T(Pyramid Pooling Transformer)

Paper	P2T: Pyramid Pooling Transformer for Scene Understanding
Date	22 Jun 2021
Code	https://github.com/yuhuan-wu/P2T
Authors	Yu-Huan Wu, Yun Liu, Xin Zhan, Ming-Ming Cheng
Org.	TKLNDST, College of Computer Science, Nankai University / Computer Vision Lab, ETH Zurich / Alibaba Group
Abstract	This paper jointly resolves two problems in vision transformer: i) the computation of Multi-Head Self-Attention (MHSA) has high computational/space complexity; ii) recent vision transformer networks are overly tuned for image classification, ignoring the difference between image classification (simple scenarios, more similar to NLP) and downstream scene understanding tasks (complicated scenarios, rich structural and contextual information). To this end, we note that pyramid pooling has been demonstrated to be effective in various vision tasks owing to its powerful context abstraction, and its natural property of spatial invariance is suitable to address the loss of structural information (problem ii)). Hence, we propose to adapt pyramid pooling to MHSA for alleviating its high requirement on computational resources (problem i)). In this way, this pooling-based MHSA can well address the above two problems and is thus flexible and powerful for downstream scene understanding tasks. Plugged with our pooling-based MHSA, we build a downstream-task-oriented transformer network, dubbed Pyramid Pooling Transformer (P2T). Extensive experiments demonstrate that, when applied P2T as the backbone network, it shows substantial superiority in various downstream scene understanding tasks such as semantic segmentation, object detection, instance segmentation, and visual saliency detection, compared to previous CNN- and transformer-based networks.

CivT(Cross inductive bias vision Transformers)

Paper	Co-advise: Cross Inductive Bias Distillation
Date	23 Jun 2021
Code
Authors	Sucheng Ren, Zhengqi Gao, Tianyu Hua, Zihui Xue, Yonglong Tian, Shengfeng He, Hang Zhao
Org.	South China University of Technology , MIT , University of Texas at Austin , Tsinghua University , Shanghai Qi Zhi Institute
Abstract	Transformers recently are adapted from the community of natural language processing as a promising substitute of convolution-based neural networks for visual learning tasks. However, its supremacy degenerates given an insufficient amount of training data (e.g., ImageNet). To make it into practical utility, we propose a novel distillation-based method to train vision transformers. Unlike previous works, where merely heavy convolution-based teachers are provided, we introduce lightweight teachers with different architectural inductive biases (e.g., convolution and involution) to co-advise the student transformer. The key is that teachers with different inductive biases attain different knowledge despite that they are trained on the same dataset, and such different knowledge compounds and boosts the student's performance during distillation. Equipped with this cross inductive bias distillation method, our vision transformers (termed as CivT) outperform all previous transformers of the same architecture on ImageNet.

VOLO(Vision Outlooker)

Paper	VOLO: Vision Outlooker for Visual Recognition
Date	24 Jun 2021
Code	https://github.com/sail-sg/volo
Authors	Li Yuan, Qibin Hou, Zihang Jiang, Jiashi Feng, Shuicheng Yan
Org.	Sea AI Lab, National University of Singapore
Abstract	Visual recognition has been dominated by convolutional neural networks (CNNs) for years. Though recently the prevailing vision transformers (ViTs) have shown great potential of self-attention based models in ImageNet classification, their performance is still inferior to that of the latest SOTA CNNs if no extra data are provided. In this work, we try to close the performance gap and demonstrate that attention-based models are indeed able to outperform CNNs. We find a major factor limiting the performance of ViTs for ImageNet classification is their low efficacy in encoding fine-level features into the token representations. To resolve this, we introduce a novel outlook attention and present a simple and general architecture, termed Vision Outlooker (VOLO). Unlike self-attention that focuses on global dependency modeling at a coarse level, the outlook attention efficiently encodes finer-level features and contexts into tokens, which is shown to be critically beneficial to recognition performance but largely ignored by the self-attention. Experiments show that our VOLO achieves 87.1% top-1 accuracy on ImageNet-1K classification, which is the first model exceeding 87% accuracy on this competitive benchmark, without using any extra training data In addition, the pre-trained VOLO transfers well to downstream tasks, such as semantic segmentation. We achieve 84.3% mIoU score on the cityscapes validation set and 54.3% on the ADE20K validation set.

ViTAS

Paper	Vision Transformer Architecture Search
Date	25 Jun 2021
Code	https://github.com/xiusu/ViTAS
Authors	Xiu Su, Shan You, Jiyang Xie, Mingkai Zheng, Fei Wang, Chen Qian, Changshui Zhang, Xiaogang Wang, Chang Xu
Org.	The University of Sydney / SenseTime Research / Beijing University / Tsinghua University / BNRist / The Chinese University of Hong Kong
Abstract	Recently, transformers have shown great superiority in solving computer vision tasks by modeling images as a sequence of manually-split patches with self-attention mechanism. However, current architectures of vision transformers (ViTs) are simply inherited from natural language processing (NLP) tasks and have not been sufficiently investigated and optimized. In this paper, we make a further step by examining the intrinsic structure of transformers for vision tasks and propose an architecture search method, dubbed ViTAS, to search for the optimal architecture with similar hardware budgets. Concretely, we design a new effective yet efficient weight sharing paradigm for ViTs, such that architectures with different token embedding, sequence size, number of heads, width, and depth can be derived from a single super-transformer. Moreover, to cater for the variance of distinct architectures, we introduce \textit{private} class token and self-attention maps in the super-transformer. In addition, to adapt the searching for different budgets, we propose to search the sampling probability of identity operation. Experimental results show that our ViTAS attains excellent results compared to existing pure transformer architectures. For example, with 1.3G FLOPs budget, our searched architecture achieves 74.7% top-1 accuracy on ImageNet and is 2.5% superior than the current baseline ViT architecture.

2021 Q3

AutoFormer

Paper	AutoFormer: Searching Transformers for Visual Recognition
Date	1 Jul 2021
Code	https://github.com/microsoft/AutoML
Authors	Minghao Chen, Houwen Peng, Jianlong Fu, Haibin Ling
Org.	Stony Brook University, Microsoft Research Asia
Abstract	Recently, pure transformer-based models have shown great potentials for vision tasks such as image classification and detection. However, the design of transformer networks is challenging. It has been observed that the depth, embedding dimension, and number of heads can largely affect the performance of vision transformers. Previous models configure these dimensions based upon manual crafting. In this work, we propose a new one-shot architecture search framework, namely AutoFormer, dedicated to vision transformer search. AutoFormer entangles the weights of different blocks in the same layers during supernet training. Benefiting from the strategy, the trained supernet allows thousands of subnets to be very well-trained. Specifically, the performance of these subnets with weights inherited from the supernet is comparable to those retrained from scratch. Besides, the searched models, which we refer to AutoFormers, surpass the recent state-of-the-arts such as ViT and DeiT. In particular, AutoFormer-tiny/small/base achieve 74.7%/81.7%/82.4% top-1 accuracy on ImageNet with 5.7M/22.9M/53.7M parameters, respectively. Lastly, we verify the transferability of AutoFormer by providing the performance on downstream benchmarks and distillation experiments.

CSWin(Cross-Shaped Windows) Transformer

Paper	CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows
Date	1 Jul 2021
Code	https://github.com/microsoft/CSWin-Transformer
Authors	Xiaoyi Dong, Jianmin Bao, Dongdong Chen, Weiming Zhang, Nenghai Yu, Lu Yuan, Dong Chen, Baining Guo
Org.	University of Science and Technology of China , Microsoft Research Asia , Microsoft Cloud + AI
Abstract	We present CSWin Transformer, an efficient and effective Transformer-based backbone for general-purpose vision tasks. A challenging issue in Transformer design is that global self-attention is very expensive to compute whereas local self-attention often limits the field of interactions of each token. To address this issue, we develop the Cross-Shaped Window self-attention mechanism for computing self-attention in the horizontal and vertical stripes in parallel that form a cross-shaped window, with each stripe obtained by splitting the input feature into stripes of equal width. We provide a detailed mathematical analysis of the effect of the stripe width and vary the stripe width for different layers of the Transformer network which achieves strong modeling capability while limiting the computation cost. We also introduce Locally-enhanced Positional Encoding (LePE), which handles the local positional information better than existing encoding schemes. LePE naturally supports arbitrary input resolutions, and is thus especially effective and friendly for downstream tasks. Incorporated with these designs and a hierarchical structure, CSWin Transformer demonstrates competitive performance on common vision tasks. Specifically, it achieves 85.4% Top-1 accuracy on ImageNet-1K without any extra training data or label, 53.9 box AP and 46.4 mask AP on the COCO detection task, and 51.7 mIOU on the ADE20K semantic segmentation task, surpassing previous state-of-the-art Swin Transformer backbone by +1.2, +2.0, +1.4, and +2.0 respectively under the similar FLOPs setting. By further pretraining on the larger dataset ImageNet-21K, we achieve 87.5% Top-1 accuracy on ImageNet-1K and state-of-the-art segmentation performance on ADE20K with 55.2 mIoU.

Focal Transformer

Paper	Focal Self-attention for Local-Global Interactions in Vision Transformers
Date	1 Jul 2021
Code	https://github.com/microsoft/Focal-Transformer
Authors	Jianwei Yang, Chunyuan Li, Pengchuan Zhang, Xiyang Dai, Bin Xiao, Lu Yuan, Jianfeng Gao
Org.	Microsoft Research at Redmond, Microsoft Cloud + AI
Abstract	Recently, Vision Transformer and its variants have shown great promise on various computer vision tasks. The ability of capturing short- and long-range visual dependencies through self-attention is arguably the main source for the success. But it also brings challenges due to quadratic computational overhead, especially for the high-resolution vision tasks (e.g., object detection). In this paper, we present focal self-attention, a new mechanism that incorporates both fine-grained local and coarse-grained global interactions. Using this new mechanism, each token attends the closest surrounding tokens at fine granularity but the tokens far away at coarse granularity, and thus can capture both short- and long-range visual dependencies efficiently and effectively. With focal self-attention, we propose a new variant of Vision Transformer models, called Focal Transformer, which achieves superior performance over the state-of-the-art vision Transformers on a range of public image classification and object detection benchmarks. In particular, our Focal Transformer models with a moderate size of 51.1M and a larger size of 89.8M achieve 83.5 and 83.8 Top-1 accuracy, respectively, on ImageNet classification at 224x224 resolution. Using Focal Transformers as the backbones, we obtain consistent and substantial improvements over the current state-of-the-art Swin Transformers for 6 different object detection methods trained with standard 1x and 3x schedules. Our largest Focal Transformer yields 58.7/58.9 box mAPs and 50.9/51.3 mask mAPs on COCO mini-val/test-dev, and 55.4 mIoU on ADE20K for semantic segmentation, creating new SoTA on three of the most challenging computer vision tasks.

ViX(Vision Xformers)

Paper	Vision Xformers: Efficient Attention for Image Classification
Date	5 Jul 2021
Code	https://github.com/pranavphoenix/VisionXformer
Authors	Pranav Jeevan, Amit Sethi
Org.	Department of Electrical Engineering, Indian Institute of Technology Bombay
Abstract	Linear attention mechanisms provide hope for overcoming the bottleneck of quadratic complexity which restricts application of transformer models in vision tasks. We modify the ViT architecture to work on longer sequence data by replacing the quadratic attention with efficient transformers like Performer, Linformer and Nyströmformer of linear complexity creating Vision X-formers (ViX). We show that ViX performs better than ViT in image classification consuming lesser computing resources. We further show that replacing the embedding linear layer by convolutional layers in ViX further increases their performance. Our test on recent visions transformer models like LeViT and Compact Convolutional Transformer (CCT) show that replacing the attention with Nyströmformer or Performer saves GPU usage and memory without deteriorating performance. Incorporating these changes can democratize transformers by making them accessible to those with limited data and computing resources.

GLiT(Global and Local Image Transformer)

Paper	GLiT: Neural Architecture Search for Global and Local Image Transformer
Date	7 Jul 2021
Code	https://github.com/bychen515/GLiT
Authors	Boyu Chen, Peixia Li, Chuming Li, Baopu Li, Lei Bai, Chen Lin, Ming Sun, Junjie yan, Wanli Ouyang
Org.	The University of Sydney, SenseTime Group Limited, BAIDU USA LLC, University of Oxford
Abstract	We introduce the first Neural Architecture Search (NAS) method to find a better transformer architecture for image recognition. Recently, transformers without CNN-based backbones are found to achieve impressive performance for image recognition. However, the transformer is designed for NLP tasks and thus could be sub-optimal when directly used for image recognition. In order to improve the visual representation ability for transformers, we propose a new search space and searching algorithm. Specifically, we introduce a locality module that models the local correlations in images explicitly with fewer computational cost. With the locality module, our search space is defined to let the search algorithm freely trade off between global and local information as well as optimizing the low-level design choice in each module. To tackle the problem caused by huge search space, a hierarchical neural architecture search method is proposed to search the optimal vision transformer from two levels separately with the evolutionary algorithm. Extensive experiments on the ImageNet dataset demonstrate that our method can find more discriminative and efficient transformer variants than the ResNet family (e.g., ResNet101) and the baseline ViT for image classification.

ViP(Visual Parser)

Paper	Visual Parser: Representing Part-whole Hierarchies with Transformers
Date	13 Jul 2021
Code	https://github.com/kevin-ssy/ViP
Authors	Shuyang Sun, Xiaoyu Yue, Song Bai, Philip Torr
Org.	University of Oxford, ByteDance AI Lab
Abstract	Human vision is able to capture the part-whole hierarchical information from the entire scene. This paper presents the Visual Parser (ViP) that explicitly constructs such a hierarchy with transformers. ViP divides visual representations into two levels, the part level and the whole level. Information of each part represents a combination of several independent vectors within the whole. To model the representations of the two levels, we first encode the information from the whole into part vectors through an attention mechanism, then decode the global information within the part vectors back into the whole representation. By iteratively parsing the two levels with the proposed encoder-decoder interaction, the model can gradually refine the features on both levels. Experimental results demonstrate that ViP can achieve very competitive performance on three major tasks e.g. classification, detection and instance segmentation. In particular, it can surpass the previous state-of-the-art CNN backbones by a large margin on object detection. The tiny model of the ViP family with 7.2× fewer parameters and 10.9× fewer FLOPS can perform comparably with the largest model ResNeXt-101-64×4d of ResNe(X)t family. Visualization results also demonstrate that the learnt parts are highly informative of the predicting class, making ViP more explainable than previous fundamental architectures.

CMT(CNNs Meet Transformers)

Paper	CMT: Convolutional Neural Networks Meet Vision Transformers
Date	13 Jul 2021
Code
Authors	Jianyuan Guo, Kai Han, Han Wu, Chang Xu, Yehui Tang, Chunjing Xu, Yunhe Wang
Org.	Huawei Technologies, University of Sydney
Abstract	Vision transformers have been successfully applied to image recognition tasks due to their ability to capture long-range dependencies within an image. However, there are still gaps in both performance and computational cost between transformers and existing convolutional neural networks (CNNs). In this paper, we aim to address this issue and develop a network that can outperform not only the canonical transformers, but also the high-performance convolutional models. We propose a new transformer based hybrid network by taking advantage of transformers to capture long-range dependencies, and of CNNs to model local features. Furthermore, we scale it to obtain a family of models, called CMTs, obtaining much better accuracy and efficiency than previous convolution and transformer based models. In particular, our CMT-S achieves 83.5% top-1 accuracy on ImageNet, while being 14x and 2x smaller on FLOPs than the existing DeiT and EfficientNet, respectively. The proposed CMT-S also generalizes well on CIFAR10 (99.2%), CIFAR100 (91.7%), Flowers (98.7%), and other challenging vision datasets such as COCO (44.3% mAP), with considerably less computational cost.

WideNet

Paper	Go Wider Instead of Deeper
Date	25 Jul 2021
Code
Authors	Fuzhao Xue, Ziji Shi, Yuxuan Lou, Yong Liu, Yang You
Org.	National University of Singapore
Abstract	The transformer has recently achieved impressive results on various tasks. To further improve the effectiveness and efficiency of the transformer, there are two trains of thought among existing works: (1) going wider by scaling to more trainable parameters; (2) going shallower by parameter sharing or model compressing along with the depth. However, larger models usually do not scale well when fewer tokens are available to train, and advanced parallelisms are required when the model is extremely large. Smaller models usually achieve inferior performance compared to the original transformer model due to the loss of representation power. In this paper, to achieve better performance with fewer trainable parameters, we propose a framework to deploy trainable parameters efficiently, by going wider instead of deeper. Specially, we scale along model width by replacing feed-forward network (FFN) with mixture-of-experts (MoE). We then share the MoE layers across transformer blocks using individual layer normalization. Such deployment plays the role to transform various semantic representations, which makes the model more parameter-efficient and effective. To evaluate our framework, we design WideNet and evaluate it on ImageNet-1K. Our best model outperforms Vision Transformer (ViT) by 1.46% with 0.72× trainable parameters. Using 0.46× and 0.13× parameters, our WideNet can still surpass ViT and ViT-MoE by 0.83% and 2.08%, respectively.

DPT(Deformable Patch-based Transformer)

Paper	DPT: Deformable Patch-based Transformer for Visual Recognition
Date	30 Jul 2021
Code	https://github.com/CASIA-IVA-Lab/DPT
Authors	Zhiyang Chen, Yousong Zhu, Chaoyang Zhao, Guosheng Hu, Wei Zeng, Jinqiao Wang, Ming Tang
Org.	Chinese Academy of Sciences, AnyVision, Peking University
Abstract	Transformer has achieved great success in computer vision, while how to split patches in an image remains a problem. Existing methods usually use a fixed-size patch embedding which might destroy the semantics of objects. To address this problem, we propose a new Deformable Patch (DePatch) module which learns to adaptively split the images into patches with different positions and scales in a data-driven way rather than using predefined fixed patches. In this way, our method can well preserve the semantics in patches. The DePatch module can work as a plug-and-play module, which can easily be incorporated into different transformers to achieve an end-to-end training. We term this DePatch-embedded transformer as Deformable Patch-based Transformer (DPT) and conduct extensive evaluations of DPT on image classification and object detection. Results show DPT can achieve 81.9% top-1 accuracy on ImageNet classification, and 43.7% box mAP with RetinaNet, 44.3% with Mask R-CNN on MSCOCO object detection.

Evo-ViT

Paper	Evo-ViT: Slow-Fast Token Evolution for Dynamic Vision Transformer
Date	3 Aug 2021
Code	https://github.com/YifanXu74/Evo-ViT
Authors	Yifan Xu, Zhijie Zhang, Mengdan Zhang, Kekai Sheng, Ke Li, Weiming Dong, Liqing Zhang, Changsheng Xu, Xing Sun
Org.	Chinese Academy of Sciences, Shanghai Jiao Tong University, Tencent Youtu Lab, University of Chinese Academy of Sciences
Abstract	Vision transformers have recently received explosive popularity, but huge computational cost is still a severe issue. Recent efficient designs for vision transformers follow two pipelines, namely, structural compression based on local spatial prior and non-structural token pruning. However, rough token pruning breaks the spatial structure that is indispensable for local spatial prior. To take advantage of both two pipelines, this work seeks to dynamically identify uninformative tokens for each instance and trim down both the training and inference complexity while maintain complete spatial structure and information flow. To achieve this goal, we propose Evo-ViT, a self-motivated slow-fast token evolution method for vision transformers. Specifically, we conduct unstructured instance-wise token selection by taking advantage of the global class attention that is unique to vision transformers. Then, we propose to update information tokens and placeholder tokens that contribute little to the final prediction with different computational properties, namely, slow-fast updating. Thanks to the slow-fast updating mechanism that guarantees information flow and spatial structure, our Evo-ViT can accelerate vanilla transformers of both flat and deep-narrow structures from the very beginning of the training process. Experimental results demonstrate that the proposed method can significantly reduce the computational costs of vision transformers while maintaining comparable performance on image classification. For example, our method accelerates DeiT-S by over 60% throughput while only sacrificing 0.4% top-1 accuracy.

PSViT(ViT with token Pooling and attention Sharing)

Paper	PSViT: Better Vision Transformer via Token Pooling and Attention Sharing
Date	7 Aug 2021
Code
Authors	Boyu Chen, Peixia Li, Baopu Li, Chuming Li, Lei Bai, Chen Lin, Ming Sun, Junjie Yan, Wanli Ouyang
Org.	The University of Sydney, BAIDU USA LLC, SenseTime Group Limited, University of Oxford
Abstract	In this paper, we observe two levels of redundancies when applying vision transformers (ViT) for image recognition. First, fixing the number of tokens through the whole network produces redundant features at the spatial level. Second, the attention maps among different transformer layers are redundant. Based on the observations above, we propose a PSViT: a ViT with token Pooling and attention Sharing to reduce the redundancy, effectively enhancing the feature representation ability, and achieving a better speed-accuracy trade-off. Specifically, in our PSViT, token pooling can be defined as the operation that decreases the number of tokens at the spatial level. Besides, attention sharing will be built between the neighboring transformer layers for reusing the attention maps having a strong correlation among adjacent layers. Then, a compact set of the possible combinations for different token pooling and attention sharing mechanisms are constructed. Based on the proposed compact set, the number of tokens in each layer and the choices of layers sharing attention can be treated as hyper-parameters that are learned from data automatically. Experimental results show that the proposed scheme can achieve up to 6.6% accuracy improvement in ImageNet classification compared with the DeiT.

Mobile-Former

Paper	Mobile-Former: Bridging MobileNet and Transformer
Date	12 Aug 2021
Code
Authors	Yinpeng Chen, Xiyang Dai, Dongdong Chen, Mengchen Liu, Xiaoyi Dong, Lu Yuan, Zicheng Liu
Org.	Microsoft, University of Science and Technology of China
Abstract	We present Mobile-Former, a parallel design of MobileNet and Transformer with a two-way bridge in between. This structure leverages the advantage of MobileNet at local processing and transformer at global interaction. And the bridge enables bidirectional fusion of local and global features. Different with recent works on vision transformer, the transformer in Mobile-Former contains very few tokens (e.g. less than 6 tokens) that are randomly initialized, resulting in low computational cost. Combining with the proposed light-weight cross attention to model the bridge, Mobile-Former is not only computationally efficient, but also has more representation power, outperforming MobileNetV3 at low FLOP regime from 25M to 500M FLOPs on ImageNet classification. For instance, it achieves 77.9% top-1 accuracy at 294M FLOPs, gaining 1.3% over MobileNetV3 but saving 17% of computations. When transferring to object detection, Mobile-Former outperforms MobileNetV3 by 8.6 AP.

ViT-ResNAS

Paper	Searching for Efficient Multi-Stage Vision Transformers
Date	1 Sep 2021
Code	https://github.com/yilunliao/vit-search
Authors	Yi-Lun Liao, Sertac Karaman, Vivienne Sze
Org.	Massachusetts Institute of Technology
Abstract	Vision Transformer (ViT) demonstrates that Transformer for natural language processing can be applied to computer vision tasks and result in comparable performance to convolutional neural networks (CNN), which have been studied and adopted in computer vision for years. This naturally raises the question of how the performance of ViT can be advanced with design techniques of CNN. To this end, we propose to incorporate two techniques and present ViT-ResNAS, an efficient multi-stage ViT architecture designed with neural architecture search (NAS). First, we propose residual spatial reduction to decrease sequence lengths for deeper layers and utilize a multi-stage architecture. When reducing lengths, we add skip connections to improve performance and stabilize training deeper networks. Second, we propose weight-sharing NAS with multi-architectural sampling. We enlarge a network and utilize its sub-networks to define a search space. A super-network covering all sub-networks is then trained for fast evaluation of their performance. To efficiently train the super-network, we propose to sample and train multiple sub-networks with one forward-backward pass. After that, evolutionary search is performed to discover high-performance network architectures. Experiments on ImageNet demonstrate that ViT-ResNAS achieves better accuracy-MACs and accuracy-throughput trade-offs than the original DeiT and other strong baselines of ViT.

UFO-ViT(Unit Force Operated Vision Transformer)

Paper	UFO-ViT: High Performance Linear Vision Transformer without Softmax
Date	29 Sep 2021
Code
Authors	Jeong-geun Song
Org.	Kakao Enterprise
Abstract	Vision transformers have become one of the most important models for computer vision tasks. While they outperform earlier convolutional networks, the complexity quadratic to N is one of the major drawbacks when using traditional self-attention algorithms. Here we propose the UFO-ViT(Unit Force Operated Vision Trnasformer), novel method to reduce the computations of self-attention by eliminating some non-linearity. Modifying few of lines from self-attention, UFO-ViT achieves linear complexity without the degradation of performance. The proposed models outperform most transformer-based models on image classification and dense prediction tasks through most capacity regime.

2021 Q4

MobileViT

Paper	MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer
Date	5 Oct 2021
Code	https://github.com/apple/ml-cvnets
Authors	Sachin Mehta, Mohammad Rastegari
Org.	Apple
Abstract	Light-weight convolutional neural networks (CNNs) are the de-facto for mobile vision tasks. Their spatial inductive biases allow them to learn representations with fewer parameters across different vision tasks. However, these networks are spatially local. To learn global representations, self-attention-based vision trans-formers (ViTs) have been adopted. Unlike CNNs, ViTs are heavy-weight. In this paper, we ask the following question: is it possible to combine the strengths of CNNs and ViTs to build a light-weight and low latency network for mobile vision tasks? Towards this end, we introduce MobileViT, a light-weight and general-purpose vision transformer for mobile devices. MobileViT presents a different perspective for the global processing of information with transformers, i.e., transformers as convolutions. Our results show that MobileViT significantly outperforms CNN- and ViT-based networks across different tasks and datasets. On the ImageNet-1k dataset, MobileViT achieves top-1 accuracy of 78.4% with about 6 million parameters, which is 3.2% and 6.2% more accurate than MobileNetv3 (CNN-based) and DeIT (ViT-based) for a similar number of parameters. On the MS-COCO object detection task, MobileViT is 5.7% more accurate than Mo-bileNetv3 for a similar number of parameters.

UniNet

Paper	UniNet: Unified Architecture Search with Convolution, Transformer, and MLP
Date	8 Oct 2021
Code
Authors	Jihao Liu, Hongsheng Li, Guanglu Song, Xin Huang, Yu Liu
Org.	SenseTime, The Chinese University of Hong Kong
Abstract	Recently, transformer and multi-layer perceptron (MLP) architectures have achieved impressive results on various vision tasks. A few works investigated manually combining those operators to design visual network architectures, and can achieve satisfactory performances to some extent. In this paper, we propose to jointly search the optimal combination of convolution, transformer, and MLP for building a series of all-operator network architectures with high performances on visual tasks. We empirically identify that the widely-used strided convolution or pooling based down-sampling modules become the performance bottlenecks when the operators are combined to form a network. To better tackle the global context captured by the transformer and MLP operators, we propose two novel context-aware down-sampling modules, which can better adapt to the global information encoded by transformer and MLP operators. To this end, we jointly search all operators and down-sampling modules in a unified search space. Notably, Our searched network UniNet (Unified Network) outperforms state-of-the-art pure convolution-based architecture, EfficientNet, and pure transformer-based architecture, Swin-Transformer, on multiple public visual benchmarks, ImageNet classification, COCO object detection, and ADE20K semantic segmentation.

SOFT(Softmax-free Transformer)

Paper	SOFT: Softmax-free Transformer with Linear Complexity
Date	22 Oct 2021
Code	https://github.com/fudan-zvg/SOFT
Authors	Jiachen Lu, Jinghan Yao, Junge Zhang, Xiatian Zhu, Hang Xu, Weiguo Gao, Chunjing Xu, Tao Xiang, Li Zhang
Org.	Fudan University, University of Surrey, Huawei Noah’s Ark Lab
Abstract	Vision transformers (ViTs) have pushed the state-of-the-art for various visual recognition tasks by patch-wise image tokenization followed by self-attention. However, the employment of self-attention modules results in a quadratic complexity in both computation and memory usage. Various attempts on approximating the self-attention computation with linear complexity have been made in Natural Language Processing. However, an in-depth analysis in this work shows that they are either theoretically flawed or empirically ineffective for visual recognition. We further identify that their limitations are rooted in keeping the softmax self-attention during approximations. Specifically, conventional self-attention is computed by normalizing the scaled dot-product between token feature vectors. Keeping this softmax operation challenges any subsequent linearization efforts. Based on this insight, for the first time, a softmax-free transformer or SOFT is proposed. To remove softmax in self-attention, Gaussian kernel function is used to replace the dot-product similarity without further normalization. This enables a full self-attention matrix to be approximated via a low-rank matrix decomposition. The robustness of the approximation is achieved by calculating its Moore-Penrose inverse using a Newton-Raphson method. Extensive experiments on ImageNet show that our SOFT significantly improves the computational efficiency of existing ViT variants. Crucially, with a linear complexity, much longer token sequences are permitted in SOFT, resulting in superior trade-off between accuracy and complexity.

Hybrid BYOL-ViT

Paper	Hybrid BYOL-ViT: Efficient approach to deal with small Datasets
Date	8 Nov 2021
Code
Authors	Safwen Naimi, Rien van Leeuwen, Wided Souidene, Slim Ben Saoud
Org.	Philips Research, University of Carthage
Abstract	Supervised learning can learn large representational spaces, which are crucial for handling difficult learning tasks. However, due to the design of the model, classical image classification approaches struggle to generalize to new problems and new situations when dealing with small datasets. In fact, supervised learning can lose the location of image features which leads to supervision collapse in very deep architectures. In this paper, we investigate how self-supervision with strong and sufficient augmentation of unlabeled data can train effectively the first layers of a neural network even better than supervised learning, with no need for millions of labeled data. The main goal is to disconnect pixel data from annotation by getting generic task-agnostic low-level features. Furthermore, we look into Vision Transformers (ViT) and show that the low-level features derived from a self-supervised architecture can improve the robustness and the overall performance of this emergent architecture. We evaluated our method on one of the smallest open-source datasets STL-10 and we obtained a significant boost of performance from 41.66% to 83.25% when inputting low-level features from a self-supervised learning architecture to the ViT instead of the raw images.

SReT(Sliced Recursive Transformer)

Paper	Sliced Recursive Transformer
Date	9 Nov 2021
Code	https://github.com/szq0214/SReT
Authors	Zhiqiang Shen, Zechun Liu, Eric Xing
Org.	CMU, MBZUAI
Abstract	We present a neat yet effective recursive operation on vision transformers that can improve parameter utilization without involving additional parameters. This is achieved by sharing weights across depth of transformer networks. The proposed method can obtain a substantial gain (~2%) simply using naïve recursive operation, requires no special or sophisticated knowledge for designing principles of networks, and introduces minimum computational overhead to the training procedure. To reduce the additional computation caused by recursive operation while maintaining the superior accuracy, we propose an approximating method through multiple sliced group self-attentions across recursive layers which can reduce the cost consumption by 10~30% with minimal performance loss. We call our model Sliced Recursive Transformer (SReT), which is compatible with a broad range of other designs for efficient vision transformers. Our best model establishes significant improvement on ImageNet over state-of-the-art methods while containing fewer parameters. The proposed sliced recursive operation allows us to build a transformer with more than 100 or even 1000 layers effortlessly under a still small size (13~15M), to avoid difficulties in optimization when the model size is too large. The flexible scalability has shown great potential for scaling up and constructing extremely deep and large dimensionality vision transformers.

Swin Transformer V2

Paper	Swin Transformer V2: Scaling Up Capacity and Resolution
Date	18 Nov 2021
Code	https://github.com/microsoft/Swin-Transformer
Authors	Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, Furu Wei, Baining Guo
Org.	Microsoft Research Asia
Abstract	We present techniques for scaling Swin Transformer up to 3 billion parameters and making it capable of training with images of up to 1,536×1,536 resolution. By scaling up capacity and resolution, Swin Transformer sets new records on four representative vision benchmarks: 84.0% top-1 accuracy on ImageNet-V2 image classification, 63.1/54.4 box/mask mAP on COCO object detection, 59.9 mIoU on ADE20K semantic segmentation, and 86.8% top-1 accuracy on Kinetics-400 video action classification. Our techniques are generally applicable for scaling up vision models, which has not been widely explored as that of NLP language models, partly due to the following difficulties in training and applications: 1) vision models often face instability issues at scale and 2) many downstream vision tasks require high resolution images or windows and it is not clear how to effectively transfer models pre-trained at low resolutions to higher resolution ones. The GPU memory consumption is also a problem when the image resolution is high. To address these issues, we present several techniques, which are illustrated by using Swin Transformer as a case study: 1) a post normalization technique and a scaled cosine attention approach to improve the stability of large vision models; 2) a log-spaced continuous position bias technique to effectively transfer models pre-trained at low-resolution images and windows to their higher-resolution counterparts. In addition, we share our crucial implementation details that lead to significant savings of GPU memory consumption and thus make it feasible to train large vision models with regular GPUs. Using these techniques and self-supervised pre-training, we successfully train a strong 3B Swin Transformer model and effectively transfer it to various vision tasks involving high-resolution images or windows, achieving the state-of-the-art accuracy on a variety of benchmarks.

SWAT(Spatial Structure Within and Among Tokens)

Paper	SWAT: Spatial Structure Within and Among Tokens
Date	26 Nov 2021
Code
Authors	Kumara Kahatapitiya, Michael S. Ryoo
Org.	Stony Brook University
Abstract	Modeling visual data as tokens (i.e., image patches), and applying attention mechanisms or feed-forward networks on top of them has shown to be highly effective in recent years. The common pipeline in such approaches includes a tokenization method, followed by a set of layers/blocks for information mixing, both within tokens and among tokens. In common practice, image patches are flattened when converted into tokens, discarding the spatial structure within each patch. Next, a module such as multi-head self-attention captures the pairwise relations among the tokens and mixes them. In this paper, we argue that models can have significant gains when spatial structure is preserved in tokenization, and is explicitly used in the mixing stage. We propose two key contributions: (1) Structure-aware Tokenization and, (2) Structure-aware Mixing, both of which can be combined with existing models with minimal effort. We introduce a family of models (SWAT), showing improvements over the likes of DeiT, MLP-Mixer and Swin Transformer, across multiple benchmarks including ImageNet classification and ADE20K segmentation. Our code and models will be released online.

SiT(Self-slimmed Vision Transformer)

Paper	Self-slimmed Vision Transformer
Date	24 Nov 2021
Code
Authors	Zhuofan Zong, Kunchang Li, Guanglu Song, Yali Wang, Yu Qiao, Biao Leng, Yu Liu
Org.	Beihang University, SenseTime Research, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Shanghai AI Laboratory
Abstract	Vision transformers (ViTs) have become the popular structures and outperformed convolutional neural networks (CNNs) on various vision tasks. However, such powerful transformers bring a huge computation burden. And the essential barrier behind this is the exhausting token-to-token comparison. To alleviate this, we delve deeply into the model properties of ViT and observe that ViTs exhibit sparse attention with high token similarity. This intuitively introduces us a feasible structure-agnostic dimension, token number, to reduce the computational cost. Based on this exploration, we propose a generic self-slimmed learning approach for vanilla ViTs, namely SiT. Specifically, we first design a novel Token Slimming Module (TSM), which can boost the inference efficiency of ViTs by dynamic token aggregation. Different from the token hard dropping, our TSM softly integrates redundant tokens into fewer informative ones, which can dynamically zoom visual attention without cutting off discriminative token relations in the images. Furthermore, we introduce a concise Dense Knowledge Distillation (DKD) framework, which densely transfers unorganized token information in a flexible auto-encoder manner. Due to the similar structure between teacher and student, our framework can effectively leverage structure knowledge for better convergence. Finally, we conduct extensive experiments to evaluate our SiT. It demonstrates that our method can speed up ViTs by 1.7x with negligible accuracy drop, and even speed up ViTs by 3.6x while maintaining 97% of their performance. Surprisingly, by simply arming LV-ViT with our SiT, we achieve new state-of-the-art performance on ImageNet, surpassing all the CNNs and ViTs in the recent literature.

S3(Searching the Search Space)

Paper	Searching the Search Space of Vision Transformer
Date	29 Nov 2021
Code	https://github.com/microsoft/Cream
Authors	Minghao Chen, Kan Wu, Bolin Ni, Houwen Peng, Bei Liu, Jianlong Fu, Hongyang Chao, Haibin Ling
Org.	Stony Brook University, Sun Yat-sen University, Institute of Automation, CAS, Microsoft Research Asia
Abstract	Vision Transformer has shown great visual representation power in substantial vision tasks such as recognition and detection, and thus been attracting fast-growing efforts on manually designing more effective architectures. In this paper, we propose to use neural architecture search to automate this process, by searching not only the architecture but also the search space. The central idea is to gradually evolve different search dimensions guided by their E-T Error computed using a weight-sharing supernet. Moreover, we provide design guidelines of general vision transformers with extensive analysis according to the space searching process, which could promote the understanding of vision transformer. Remarkably, the searched models, named S3 (short for Searching the Search Space), from the searched space achieve superior performance to recently proposed models, such as Swin, DeiT and ViT, when evaluated on ImageNet. The effectiveness of S3 is also illustrated on object detection, semantic segmentation and visual question answering, demonstrating its generality to downstream vision and vision-language tasks.

AdaViT(Adaptive Vision Transformers)

Paper	AdaViT: Adaptive Vision Transformers for Efficient Image Recognition
Date	30 Nov 2021
Code
Authors	Lingchen Meng, Hengduo Li, Bor-Chun Chen, Shiyi Lan, Zuxuan Wu, Yu-Gang Jiang, Ser-Nam Lim
Org.	Fudan University, University of Maryland, Meta AI
Abstract	Built on top of self-attention mechanisms, vision transformers have demonstrated remarkable performance on a variety of vision tasks recently. While achieving excellent performance, they still require relatively intensive computational cost that scales up drastically as the numbers of patches, self-attention heads and transformer blocks increase. In this paper, we argue that due to the large variations among images, their need for modeling long-range dependencies between patches differ. To this end, we introduce AdaViT, an adaptive computation framework that learns to derive usage policies on which patches, self-attention heads and transformer blocks to use throughout the backbone on a per-input basis, aiming to improve inference efficiency of vision transformers with a minimal drop of accuracy for image recognition. Optimized jointly with a transformer backbone in an end-to-end manner, a light-weight decision network is attached to the backbone to produce decisions on-the-fly. Extensive experiments on ImageNet demonstrate that our method obtains more than 2x improvement on efficiency compared to state-of-the-art vision transformers with only 0.8% drop of accuracy, achieving good efficiency/accuracy trade-offs conditioned on different computational budgets. We further conduct quantitative and qualitative analysis on learned usage polices and provide more insights on the redundancy in vision transformers.

Shunted-Transformer

Paper	Shunted Self-Attention via Multi-Scale Token Aggregation
Date	30 Nov 2021
Code	https://github.com/OliverRensu/Shunted-Transformer
Authors	Sucheng Ren, Daquan Zhou, Shengfeng He, Jiashi Feng, Xinchao Wang
Org.	South China University of Technology, National University of Singapore, Sea AI Lab
Abstract	Recent Vision Transformer~(ViT) models have demonstrated encouraging results across various computer vision tasks, thanks to their competence in modeling long-range dependencies of image patches or tokens via self-attention. These models, however, usually designate the similar receptive fields of each token feature within each layer. Such a constraint inevitably limits the ability of each self-attention layer in capturing multi-scale features, thereby leading to performance degradation in handling images with multiple objects of different scales. To address this issue, we propose a novel and generic strategy, termed shunted self-attention~(SSA), that allows ViTs to model the attentions at hybrid scales per attention layer. The key idea of SSA is to inject heterogeneous receptive field sizes into tokens: before computing the self-attention matrix, it selectively merges tokens to represent larger object features while keeping certain tokens to preserve fine-grained features. This novel merging scheme enables the self-attention to learn relationships between objects with different sizes and simultaneously reduces the token numbers and the computational cost. Extensive experiments across various tasks demonstrate the superiority of SSA. Specifically, the SSA-based transformer achieves 84.0% Top-1 accuracy and outperforms the state-of-the-art Focal Transformer on ImageNet with only half of the model size and computation cost, and surpasses Focal Transformer by 1.3 mAP on COCO and 2.9 mIOU on ADE20K under similar parameter and computation cost.

MViTv2(Multiscale Vision Transformers)

Paper	Improved Multiscale Vision Transformers for Classification and Detection
Date	2 Dec 2021
Code	https://github.com/facebookresearch/SlowFast
Authors	Yanghao Li, Chao-Yuan Wu, Haoqi Fan, Karttikeya Mangalam, Bo Xiong, Jitendra Malik, Christoph Feichtenhofer
Org.	Facebook AI Research , UC Berkeley
Abstract	In this paper, we study Multiscale Vision Transformers (MViT) as a unified architecture for image and video classification, as well as object detection. We present an improved version of MViT that incorporates decomposed relative positional embeddings and residual pooling connections. We instantiate this architecture in five sizes and evaluate it for ImageNet classification, COCO detection and Kinetics video recognition where it outperforms prior work. We further compare MViTs' pooling attention to window attention mechanisms where it outperforms the latter in accuracy/compute. Without bells-and-whistles, MViT has state-of-the-art performance in 3 domains: 88.8% accuracy on ImageNet classification, 56.1 box AP on COCO object detection as well as 86.1% on Kinetics-400 video classification.

ReViT(Resizable-ViT)

Paper	Make A Long Image Short: Adaptive Token Length for Vision Transformers
Date	3 Dec 2021
Code
Authors	Yichen Zhu, Yuqin Zhu, Jie Du, Yi Wang, Zhicai Ou, Feifei Feng, Jian Tang
Org.	Midea Group, AI Innovation Center / Tongji University
Abstract	The vision transformer splits each image into a sequence of tokens with fixed length and processes the tokens in the same way as words in natural language processing. More tokens normally lead to better performance but considerably increased computational cost. Motivated by the proverb "A picture is worth a thousand words" we aim to accelerate the ViT model by making a long image short. To this end, we propose a novel approach to assign token length adaptively during inference. Specifically, we first train a ViT model, called Resizable-ViT (ReViT), that can process any given input with diverse token lengths. Then, we retrieve the "token-length label" from ReViT and use it to train a lightweight Token-Length Assigner (TLA). The token-length labels are the smallest number of tokens to split an image that the ReViT can make the correct prediction, and TLA is learned to allocate the optimal token length based on these labels. The TLA enables the ReViT to process the image with the minimum sufficient number of tokens during inference. Thus, the inference speed is boosted by reducing the token numbers in the ViT model. Our approach is general and compatible with modern vision transformer architectures and can significantly reduce computational expanse. We verified the effectiveness of our methods on multiple representative ViT models (DeiT, LV-ViT, and TimesFormer) across two tasks (image classification and action recognition).

Locally-SAG-Vit

Paper	Locally Shifted Attention With Early Global Integration
Date	9 Dec 2021
Code	https://github.com/shellysheynin/Locally-SAG-Transformer
Authors	Shelly Sheynin, Sagie Benaim, Adam Polyak, Lior Wolf
Org.	Tel Aviv University / Facebook AI Research
Abstract	Recent work has shown the potential of transformers for computer vision applications. An image is first partitioned into patches, which are then used as input tokens for the attention mechanism. Due to the expensive quadratic cost of the attention mechanism, either a large patch size is used, resulting in coarse-grained global interactions, or alternatively, attention is applied only on a local region of the image, at the expense of long-range interactions. In this work, we propose an approach that allows for both coarse global interactions and fine-grained local interactions already at early layers of a vision transformer.At the core of our method is the application of local and global attention layers. In the local attention layer, we apply attention to each patch and its local shifts, resulting in virtually located local patches, which are not bound to a single, specific location. These virtually located patches are then used in a global attention layer. The separation of the attention layer into local and global counterparts allows for a low computational cost in the number of patches, while still supporting data-dependent localization already at the first layer, as opposed to the static positioning in other visual transformers. Our method is shown to be superior to both convolutional and transformer-based methods for image classification on CIFAR10, CIFAR100, and ImageNet.

AdaViT

Paper	AdaViT: Adaptive Tokens for Efficient Vision Transformer
Date	14 Dec 2021
Code
Authors	Hongxu Yin, Arash Vahdat, Jose Alvarez, Arun Mallya, Jan Kautz, Pavlo Molchanov
Org.	NVIDIA
Abstract	We introduce AdaViT, a method that adaptively adjusts the inference cost of vision transformer (ViT) for images of different complexity. AdaViT achieves this by automatically reducing the number of tokens in vision transformers that are processed in the network as inference proceeds. We reformulate Adaptive Computation Time (ACT) for this task, extending halting to discard redundant spatial tokens. The appealing architectural properties of vision transformers enables our adaptive token reduction mechanism to speed up inference without modifying the network architecture or inference hardware. We demonstrate that AdaViT requires no extra parameters or sub-network for halting, as we base the learning of adaptive halting on the original network parameters. We further introduce distributional prior regularization that stabilizes training compared to prior ACT approaches. On the image classification task (ImageNet1K), we show that our proposed AdaViT yields high efficacy in filtering informative spatial features and cutting down on the overall compute. The proposed method improves the throughput of DeiT-Tiny by 62% and DeiT-Small by 38% with only 0.3% accuracy drop, outperforming prior art by a large margin.

LVT(Lite Vision Transformer)

Paper	Lite Vision Transformer with Enhanced Self-Attention
Date	20 Dec 2021
Code	https://github.com/Chenglin-Yang/LVT
Authors	Chenglin Yang, Yilin Wang, Jianming Zhang, He Zhang, Zijun Wei, Zhe Lin, Alan Yuille
Org.	Johns Hopkins University, Adobe Inc.
Abstract	Despite the impressive representation capacity of vision transformer models, current light-weight vision transformer models still suffer from inconsistent and incorrect dense predictions at local regions. We suspect that the power of their self-attention mechanism is limited in shallower and thinner networks. We propose Lite Vision Transformer (LVT), a novel light-weight transformer network with two enhanced self-attention mechanisms to improve the model performances for mobile deployment. For the low-level features, we introduce Convolutional Self-Attention (CSA). Unlike previous approaches of merging convolution and self-attention, CSA introduces local self-attention into the convolution within a kernel of size 3x3 to enrich low-level features in the first stage of LVT. For the high-level features, we propose Recursive Atrous Self-Attention (RASA), which utilizes the multi-scale context when calculating the similarity map and a recursive mechanism to increase the representation capability with marginal extra parameter cost. The superiority of LVT is demonstrated on ImageNet recognition, ADE20K semantic segmentation, and COCO panoptic segmentation.

ELSA

Paper	ELSA: Enhanced Local Self-Attention for Vision Transformer
Date	23 Dec 2021
Code	https://github.com/damo-cv/ELSA
Authors	Jingkai Zhou, Pichao Wang, Fan Wang, Qiong Liu, Hao Li, Rong Jin
Org.	South China University of Technology , Alibaba Group
Abstract	Self-attention is powerful in modeling long-range dependencies, but it is weak in local finer-level feature learning. The performance of local self-attention (LSA) is just on par with convolution and inferior to dynamic filters, which puzzles researchers on whether to use LSA or its counterparts, which one is better, and what makes LSA mediocre. To clarify these, we comprehensively investigate LSA and its counterparts from two sides: \emph{channel setting} and \emph{spatial processing}. We find that the devil lies in the generation and application of spatial attention, where relative position embeddings and the neighboring filter application are key factors. Based on these findings, we propose the enhanced local self-attention (ELSA) with Hadamard attention and the ghost head. Hadamard attention introduces the Hadamard product to efficiently generate attention in the neighboring case, while maintaining the high-order mapping. The ghost head combines attention maps with static matrices to increase channel capacity. Experiments demonstrate the effectiveness of ELSA. Without architecture / hyperparameter modification, drop-in replacing LSA with ELSA boosts Swin Transformer \cite{swin} by up to +1.4 on top-1 accuracy. ELSA also consistently benefits VOLO \cite{volo} from D1 to D5, where ELSA-VOLO-D5 achieves 87.2 on the ImageNet-1K without extra training images. In addition, we evaluate ELSA in downstream tasks. ELSA significantly improves the baseline by up to +1.9 box Ap / +1.3 mask Ap on the COCO, and by up to +1.9 mIoU on the ADE20K.

SimViT(Simple ViT)

Paper	SimViT: Exploring a Simple Vision Transformer with sliding windows
Date	24 Dec 2021
Code	https://github.com/ucasligang/SimViT
Authors	Gang Li, Di Xu, Xing Cheng, Lingyu Si, Changwen Zheng
Org.	Chinese Academy of Science
Abstract	Although vision Transformers have achieved excellent performance as backbone models in many vision tasks, most of them intend to capture global relations of all tokens in an image or a window, which disrupts the inherent spatial and local correlations between patches in 2D structure. In this paper, we introduce a simple vision Transformer named SimViT, to incorporate spatial structure and local information into the vision Transformers. Specifically, we introduce Multi-head Central Self-Attention(MCSA) instead of conventional Multi-head Self-Attention to capture highly local relations. The introduction of sliding windows facilitates the capture of spatial structure. Meanwhile, SimViT extracts multi-scale hierarchical features from different layers for dense prediction tasks. Extensive experiments show the SimViT is effective and efficient as a general-purpose backbone model for various image processing tasks. Especially, our SimViT-Micro only needs 3.3M parameters to achieve 71.1% top-1 accuracy on ImageNet-1k dataset, which is the smallest size vision Transformer model by now.

SPT and LSA

Paper	Vision Transformer for Small-Size Datasets
Date	27 Dec 2021
Code	https://github.com/aanna0701/SPT_LSA_ViT
Authors	Seung Hoon Lee, Seunghyun Lee, Byung Cheol Song
Org.	Inha University
Abstract	Recently, the Vision Transformer (ViT), which applied the transformer structure to the image classification task, has outperformed convolutional neural networks. However, the high performance of the ViT results from pre-training using a large-size dataset such as JFT-300M, and its dependence on a large dataset is interpreted as due to low locality inductive bias. This paper proposes Shifted Patch Tokenization (SPT) and Locality Self-Attention (LSA), which effectively solve the lack of locality inductive bias and enable it to learn from scratch even on small-size datasets. Moreover, SPT and LSA are generic and effective add-on modules that are easily applicable to various ViTs. Experimental results show that when both SPT and LSA were applied to the ViTs, the performance improved by an average of 2.96% in Tiny-ImageNet, which is a representative small-size dataset. Especially, Swin Transformer achieved an overwhelming performance improvement of 4.08% thanks to the proposed SPT and LSA.

SPVit(Soft Token Pruning ViT)

Paper	SPViT: Enabling Faster Vision Transformers via Soft Token Pruning
Date	27 Dec 2021
Code
Authors	Zhenglun Kong, Peiyan Dong, Xiaolong Ma, Xin Meng, Wei Niu, Mengshu Sun, Bin Ren, Minghai Qin, Hao Tang, Yanzhi Wang
Org.	Northeastern University , College of William and Mary , Peking University , CVL, CoCoPIE LLC
Abstract	Recently, Vision Transformer (ViT) has continuously established new milestones in the computer vision field, while the high computation and memory cost makes its propagation in industrial production difficult. Pruning, a traditional model compression paradigm for hardware efficiency, has been widely applied in various DNN structures. Nevertheless, it stays ambiguous on how to perform exclusive pruning on the ViT structure. Considering three key points: the structural characteristics, the internal data pattern of ViTs, and the related edge device deployment, we leverage the input token sparsity and propose a computation-aware soft pruning framework, which can be set up on vanilla Transformers of both flatten and CNN-type structures, such as Pooling-based ViT (PiT). More concretely, we design a dynamic attention-based multi-head token selector, which is a lightweight module for adaptive instance-wise token selection. We further introduce a soft pruning technique, which integrates the less informative tokens generated by the selector module into a package token that will participate in subsequent calculations rather than being completely discarded. Our framework is bound to the trade-off between accuracy and computation constraints of specific edge devices through our proposed computation-aware training strategy. Experimental results show that our framework significantly reduces the computation cost of ViTs while maintaining comparable performance on image classification. Moreover, our framework can guarantee the identified model to meet resource specifications of mobile devices and FPGA, and even achieve the real-time execution of DeiT-T on mobile platforms. For example, our method reduces the latency of DeiT-T to 26 ms (26%∼41% superior to existing works) on the mobile device with 0.25%∼4% higher top-1 accuracy on ImageNet.

Pale Transformer

Paper	Pale Transformer: A General Vision Transformer Backbone with Pale-Shaped Attention
Date	28 Dec 2021
Code	https://github.com/BR-IDL/PaddleViT
Authors	Sitong Wu, Tianyi Wu, Haoru Tan, Guodong Guo
Org.	Baidu Research , National Engineering Laboratory for Deep Learning Technology and Application, University of Chinese Academy of Sciences
Abstract	Recently, Transformers have shown promising performance in various vision tasks. To reduce the quadratic computation complexity caused by the global self-attention, various methods constrain the range of attention within a local region to improve its efficiency. Consequently, their receptive fields in a single attention layer are not large enough, resulting in insufficient context modeling. To address this issue, we propose a Pale-Shaped self-Attention (PS-Attention), which performs self-attention within a pale-shaped region. Compared to the global self-attention, PS-Attention can reduce the computation and memory costs significantly. Meanwhile, it can capture richer contextual information under the similar computation complexity with previous local self-attention mechanisms. Based on the PS-Attention, we develop a general Vision Transformer backbone with a hierarchical architecture, named Pale Transformer, which achieves 83.4%, 84.3%, and 84.9% Top-1 accuracy with the model size of 22M, 48M, and 85M respectively for 224 ImageNet-1K classification, outperforming the previous Vision Transformer backbones. For downstream tasks, our Pale Transformer backbone performs better than the recent state-of-the-art CSWin Transformer by a large margin on ADE20K semantic segmentation and COCO object detection & instance segmentation.

2022 Q1

DAT(Deformable Attention Transformer)

Paper	Vision Transformer with Deformable Attention
Date	3 Jan 2022
Code	https://github.com/LeapLabTHU/DAT
Authors	Zhuofan Xia, Xuran Pan, Shiji Song, Li Erran Li, Gao Huang
Org.	Tsinghua University , Amazon , Beijing Academy of Artificial Intelligence
Abstract	Transformers have recently shown superior performances on various vision tasks. The large, sometimes even global, receptive field endows Transformer models with higher representation power over their CNN counterparts. Nevertheless, simply enlarging receptive field also gives rise to several concerns. On the one hand, using dense attention e.g., in ViT, leads to excessive memory and computational cost, and features can be influenced by irrelevant parts which are beyond the region of interests. On the other hand, the sparse attention adopted in PVT or Swin Transformer is data agnostic and may limit the ability to model long range relations. To mitigate these issues, we propose a novel deformable self-attention module, where the positions of key and value pairs in self-attention are selected in a data-dependent way. This flexible scheme enables the self-attention module to focus on relevant regions and capture more informative features. On this basis, we present Deformable Attention Transformer, a general backbone model with deformable attention for both image classification and dense prediction tasks. Extensive experiments show that our models achieve consistently improved results on comprehensive benchmarks.

ViT-Slim

Paper	Vision Transformer Slimming: Multi-Dimension Searching in Continuous Optimization Space
Date	3 Jan 2022
Code	https://github.com/arnav0400/vit-slim
Authors	Arnav Chavan, Zhiqiang Shen, Zhuang Liu, Zechun Liu, Kwang-Ting Cheng, Eric Xing
Org.	IIT Dhanbad ,CMU ,MBZUAI ,HKUST,UC Berkeley
Abstract	This paper explores the feasibility of finding an optimal sub-model from a vision transformer and introduces a pure vision transformer slimming (ViT-Slim) framework that can search such a sub-structure from the original model end-to-end across multiple dimensions, including the input tokens, MHSA and MLP modules with state-of-the-art performance. Our method is based on a learnable and unified l1 sparsity constraint with pre-defined factors to reflect the global importance in the continuous searching space of different dimensions. The searching process is highly efficient through a single-shot training scheme. For instance, on DeiT-S, ViT-Slim only takes ~43 GPU hours for searching process, and the searched structure is flexible with diverse dimensionalities in different modules. Then, a budget threshold is employed according to the requirements of accuracy-FLOPs trade-off on running devices, and a re-training process is performed to obtain the final models. The extensive experiments show that our ViT-Slim can compress up to 40% of parameters and 40% FLOPs on various vision transformers while increasing the accuracy by ~0.6% on ImageNet. We also demonstrate the advantage of our searched models on several downstream datasets.

PyramidTNT(Pyramid Transformer-in-Transformer)

Paper	PyramidTNT: Improved Transformer-in-Transformer Baselines with Pyramid Architecture
Date	4 Jan 2022
Code	https://github.com/huawei-noah/CV-Backbones/tree/master/tnt_pytorch
Authors	Kai Han, Jianyuan Guo, Yehui Tang, Yunhe Wang
Org.	Huawei Noah’s Ark Lab
Abstract	Transformer networks have achieved great progress for computer vision tasks. Transformer-in-Transformer (TNT) architecture utilizes inner transformer and outer transformer to extract both local and global representations. In this work, we present new TNT baselines by introducing two advanced designs: 1) pyramid architecture, and 2) convolutional stem. The new "PyramidTNT" significantly improves the original TNT by establishing hierarchical representations. PyramidTNT achieves better performances than the previous state-of-the-art vision transformers such as Swin Transformer. We hope this new baseline will be helpful to the further research and application of vision transformer.

QuadTree

Paper	QuadTree Attention for Vision Transformers
Date	8 Jan 2022
Code	https://github.com/Tangshitao/QuadtreeAttention
Authors	Shitao Tang, Jiahui Zhang, Siyu Zhu, Ping Tan
Org.	Simon Fraser University, Alibaba A.I. Lab
Abstract	Transformers have been successful in many vision tasks, thanks to their capability of capturing long-range dependency. However, their quadratic computational complexity poses a major obstacle for applying them to vision tasks requiring dense predictions, such as object detection, feature matching, stereo, etc. We introduce QuadTree Attention, which reduces the computational complexity from quadratic to linear. Our quadtree transformer builds token pyramids and computes attention in a coarse-to-fine manner. At each level, the top K patches with the highest attention scores are selected, such that at the next level, attention is only evaluated within the relevant regions corresponding to these top K patches. We demonstrate that quadtree attention achieves state-of-the-art performance in various vision tasks, e.g. with 4.0% improvement in feature matching on ScanNet, about 50% flops reduction in stereo matching, 0.4-1.5% improvement in top-1 accuracy on ImageNet classification, 1.2-1.8% improvement on COCO object detection, and 0.7-2.4% improvement on semantic segmentation over previous state-of-the-art transformers.

UniFormer

Paper	UniFormer: Unifying Convolution and Self-attention for Visual Recognition
Date	24 Jan 2022
Code	https://github.com/Sense-X/UniFormer
Authors	Kunchang Li, Yali Wang, Junhao Zhang, Peng Gao, Guanglu Song, Yu Liu, Hongsheng Li, Yu Qiao
Org.	Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, National University of Singapore, Shanghai AI Laboratory, Chinese University of Hong Kong
Abstract	It is a challenging task to learn discriminative representation from images and videos, due to large local redundancy and complex global dependency in these visual data. Convolution neural networks (CNNs) and vision transformers (ViTs) have been two dominant frameworks in the past few years. Though CNNs can efficiently decrease local redundancy by convolution within a small neighborhood, the limited receptive field makes it hard to capture global dependency. Alternatively, ViTs can effectively capture long-range dependency via self-attention, while blind similarity comparisons among all the tokens lead to high redundancy. To resolve these problems, we propose a novel Unified transFormer (UniFormer), which can seamlessly integrate the merits of convolution and self-attention in a concise transformer format. Different from the typical transformer blocks, the relation aggregators in our UniFormer block are equipped with local and global token affinity respectively in shallow and deep layers, allowing to tackle both redundancy and dependency for efficient and effective representation learning. Finally, we flexibly stack our UniFormer blocks into a new powerful backbone, and adopt it for various vision tasks from image to video domain, from classification to dense prediction. Without any extra training data, our UniFormer achieves 86.3 top-1 accuracy on ImageNet-1K classification. With only ImageNet-1K pre-training, it can simply achieve state-of-the-art performance in a broad range of downstream tasks, e.g., it obtains 82.9/84.8 top-1 accuracy on Kinetics-400/600, 60.9/71.2 top-1 accuracy on Something-Something V1/V2 video classification tasks, 53.8 box AP and 46.4 mask AP on COCO object detection task, 50.8 mIoU on ADE20K semantic segmentation task, and 77.4 AP on COCO pose estimation task.

CXV(Convolutional Xformers for Vision)

Paper	Convolutional Xformers for Vision
Date	25 Jan 2022
Code	https://github.com/pranavphoenix/CXV
Authors	Pranav Jeevan, Amit sethi
Org.	Indian Institute of Technology Bombay
Abstract	Vision transformers (ViTs) have found only limited practical use in processing images, in spite of their state-of-the-art accuracy on certain benchmarks. The reason for their limited use include their need for larger training datasets and more computational resources compared to convolutional neural networks (CNNs), owing to the quadratic complexity of their self-attention mechanism. We propose a linear attention-convolution hybrid architecture -- Convolutional X-formers for Vision (CXV) -- to overcome these limitations. We replace the quadratic attention with linear attention mechanisms, such as Performer, Nyströmformer, and Linear Transformer, to reduce its GPU usage. Inductive prior for image data is provided by convolutional sub-layers, thereby eliminating the need for class token and positional embeddings used by the ViTs. We also propose a new training method where we use two different optimizers during different phases of training and show that it improves the top-1 image classification accuracy across different architectures. CXV outperforms other architectures, token mixers (e.g. ConvMixer, FNet and MLP Mixer), transformer models (e.g. ViT, CCT, CvT and hybrid Xformers), and ResNets for image classification in scenarios with limited data and GPU resources (cores, RAM, power).

O-ViT(Orthogonal Vision Transformer)

Paper	O-ViT: Orthogonal Vision Transformer
Date	28 Jan 2022
Code
Authors	Yanhong Fei, Yingjie Liu, Xian Wei, Mingsong Chen
Org.	East China Normal University, Chinese Academy of Sciences
Abstract	Inspired by the tremendous success of the self-attention mechanism in natural language processing, the Vision Transformer (ViT) creatively applies it to image patch sequences and achieves incredible performance. However, the scaled dot-product self-attention of ViT brings about scale ambiguity to the structure of the original feature space. To address this problem, we propose a novel method named Orthogonal Vision Transformer (O-ViT), to optimize ViT from the geometric perspective. O-ViT limits parameters of self-attention blocks to be on the norm-keeping orthogonal manifold, which can keep the geometry of the feature space. Moreover, O-ViT achieves both orthogonal constraints and cheap optimization overhead by adopting a surjective mapping between the orthogonal group and its Lie algebra.We have conducted comparative experiments on image recognition tasks to demonstrate O-ViT's validity and experiments show that O-ViT can boost the performance of ViT by up to 3.6%.

MOA(Multi-resolution Overlapped Attention)

Paper	Aggregating Global Features into Local Vision Transformer
Date	30 Jan 2022
Code	https://github.com/krushi1992/moa-transformer
Authors	Krushi Patel, Andres M. Bur, Fengjun Li, Guanghui Wang
Org.	University of Kansas, Ryerson University
Abstract	Local Transformer-based classification models have recently achieved promising results with relatively low computational costs. However, the effect of aggregating spatial global information of local Transformer-based architecture is not clear. This work investigates the outcome of applying a global attention-based module named multi-resolution overlapped attention (MOA) in the local window-based transformer after each stage. The proposed MOA employs slightly larger and overlapped patches in the key to enable neighborhood pixel information transmission, which leads to significant performance gain. In addition, we thoroughly investigate the effect of the dimension of essential architecture components through extensive experiments and discover an optimum architecture design. Extensive experimental results CIFAR-10, CIFAR-100, and ImageNet-1K datasets demonstrate that the proposed approach outperforms previous vision Transformers with a comparatively fewer number of parameters.

BOAT(Bilateral Local Attention Vision Transformer)

Paper	BOAT: Bilateral Local Attention Vision Transformer
Date	31 Jan 2022
Code
Authors	Tan Yu, Gangming Zhao, Ping Li, Yizhou Yu
Org.	Baidu Research, The University of Hong Kong
Abstract	Vision Transformers achieved outstanding performance in many computer vision tasks. Early Vision Transformers such as ViT and DeiT adopt global self-attention, which is computationally expensive when the number of patches is large. To improve efficiency, recent Vision Transformers adopt local self-attention mechanisms, where self-attention is computed within local windows. Despite the fact that window-based local self-attention significantly boosts efficiency, it fails to capture the relationships between distant but similar patches in the image plane. To overcome this limitation of image-space local attention, in this paper, we further exploit the locality of patches in the feature space. We group the patches into multiple clusters using their features, and self-attention is computed within every cluster. Such feature-space local attention effectively captures the connections between patches across different local windows but still relevant. We propose a Bilateral lOcal Attention vision Transformer (BOAT), which integrates feature-space local attention with image-space local attention. We further integrate BOAT with both Swin and CSWin models, and extensive experiments on several benchmark datasets demonstrate that our BOAT-CSWin model clearly and consistently outperforms existing state-of-the-art CNN models and vision Transformers.

BViT(Broad Attention based Vision Transformer)

Paper	BViT: Broad Attention based Vision Transformer
Date	13 Feb 2022
Code	https://github.com/DRL-CASIA/Broad_ViT
Authors	Nannan Li, Yaran Chen, Weifan Li, Zixiang Ding, Dongbin Zhao
Org.	Chinese Academy of Sciences
Abstract	Recent works have demonstrated that transformer can achieve promising performance in computer vision, by exploiting the relationship among image patches with self-attention. While they only consider the attention in a single feature layer, but ignore the complementarity of attention in different levels. In this paper, we propose the broad attention to improve the performance by incorporating the attention relationship of different layers for vision transformer, which is called BViT. The broad attention is implemented by broad connection and parameter-free attention. Broad connection of each transformer layer promotes the transmission and integration of information for BViT. Without introducing additional trainable parameters, parameter-free attention jointly focuses on the already available attention information in different layers for extracting useful information and building their relationship. Experiments on image classification tasks demonstrate that BViT delivers state-of-the-art accuracy of 74.8%/81.6% top-1 accuracy on ImageNet with 5M/22M parameters. Moreover, we transfer BViT to downstream object recognition benchmarks to achieve 98.9% and 89.9% on CIFAR10 and CIFAR100 respectively that exceed ViT with fewer parameters. For the generalization test, the broad attention in Swin Transformer and T2T-ViT also bring an improvement of more than 1%. To sum up, broad attention is promising to promote the performance of attention based models.

AlterNet

Paper	How Do Vision Transformers Work?
Date	14 Feb 2022
Code	https://github.com/xxxnell/how-do-vits-work
Authors	Namuk Park, Songkuk Kim
Org.	Yonsei University, NAVER AI Lab
Abstract	The success of multi-head self-attentions (MSAs) for computer vision is now indisputable. However, little is known about how MSAs work. We present fundamental explanations to help better understand the nature of MSAs. In particular, we demonstrate the following properties of MSAs and Vision Transformers (ViTs): (1) MSAs improve not only accuracy but also generalization by flattening the loss landscapes. Such improvement is primarily attributable to their data specificity, not long-range dependency. On the other hand, ViTs suffer from non-convex losses. Large datasets and loss landscape smoothing methods alleviate this problem; (2) MSAs and Convs exhibit opposite behaviors. For example, MSAs are low-pass filters, but Convs are high-pass filters. Therefore, MSAs and Convs are complementary; (3) Multi-stage neural networks behave like a series connection of small individual models. In addition, MSAs at the end of a stage play a key role in prediction. Based on these insights, we propose AlterNet, a model in which Conv blocks at the end of a stage are replaced with MSA blocks. AlterNet outperforms CNNs not only in large data regimes but also in small data regimes.

VAT (Visual Attention Network)

Paper	Visual Attention Network
Date	20 Feb 2022
Code	https://github.com/Visual-Attention-Network
Authors	Meng-Hao Guo, Cheng-Ze Lu, Zheng-Ning Liu, Ming-Ming Cheng, Shi-Min Hu
Org.	Tsinghua University, Nankai University
Abstract	While originally designed for natural language processing (NLP) tasks, the self-attention mechanism has recently taken various computer vision areas by storm. However, the 2D nature of images brings three challenges for applying self-attention in computer vision. (1) Treating images as 1D sequences neglects their 2D structures. (2) The quadratic complexity is too expensive for high-resolution images. (3) It only captures spatial adaptability but ignores channel adaptability. In this paper, we propose a novel large kernel attention (LKA) module to enable self-adaptive and long-range correlations in self-attention while avoiding the above issues. We further introduce a novel neural network based on LKA, namely Visual Attention Network (VAN). While extremely simple and efficient, VAN outperforms the state-of-the-art vision transformers and convolutional neural networks with a large margin in extensive experiments, including image classification, object detection, semantic segmentation, instance segmentation, etc.

ViTAEv2 (Vision Transformer Advanced by Exploring Inductive Bias v2)

Paper	ViTAEv2: Vision Transformer Advanced by Exploring Inductive Bias for Image Recognition and Beyond
Date	21 Feb 2022
Code	https://github.com/Annbless/ViTAE
Authors	Qiming Zhang, Yufei Xu, Jing Zhang, Dacheng Tao
Org.	The University of Sydney, JD Explore Academy
Abstract	Vision transformers have shown great potential in various computer vision tasks owing to their strong capability to model long-range dependency using the self-attention mechanism. Nevertheless, they treat an image as a 1D sequence of visual tokens, lacking an intrinsic inductive bias (IB) in modeling local visual structures and dealing with scale variance, which is instead learned implicitly from large-scale training data with longer training schedules. In this paper, we propose a Vision Transformer Advanced by Exploring intrinsic IB from convolutions, i.e., ViTAE. Technically, ViTAE has several spatial pyramid reduction modules to downsample and embed the input image into tokens with rich multi-scale context using multiple convolutions with different dilation rates. In this way, it acquires an intrinsic scale invariance IB and can learn robust feature representation for objects at various scales. Moreover, in each transformer layer, ViTAE has a convolution block parallel to the multi-head self-attention module, whose features are fused and fed into the feed-forward network. Consequently, it has the intrinsic locality IB and is able to learn local features and global dependencies collaboratively. The proposed two kinds of cells are stacked in both isotropic and multi-stage manners to formulate two families of ViTAE models, i.e., the vanilla ViTAE and ViTAEv2. Experiments on the ImageNet dataset as well as downstream tasks on the MS COCO, ADE20K, and AP10K datasets validate the superiority of our models over the baseline transformer models and concurrent works. Besides, we scale up our ViTAE model to 644M parameters and obtain the state-of-the-art classification performance, i.e., 88.5% Top-1 classification accuracy on ImageNet validation set and the best 91.2% Top-1 accuracy on ImageNet real validation set, without using extra private data.

PatchMerger

Paper	Learning to Merge Tokens in Vision Transformers
Date	24 Feb 2022
Code
Authors	Cedric Renggli, André Susano Pinto, Neil Houlsby, Basil Mustafa, Joan Puigcerver, Carlos Riquelme
Org.	ETH Zurich, Google Research
Abstract	Transformers are widely applied to solve natural language understanding and computer vision tasks. While scaling up these architectures leads to improved performance, it often comes at the expense of much higher computational costs. In order for large-scale models to remain practical in real-world systems, there is a need for reducing their computational overhead. In this work, we present the PatchMerger, a simple module that reduces the number of patches or tokens the network has to process by merging them between two consecutive intermediate layers. We show that the PatchMerger achieves a significant speedup across various model sizes while matching the original performance both upstream and downstream after fine-tuning.

As-ViT(Auto-scaling Vision Transformers)

Paper	Auto-scaling Vision Transformers without Training
Date	24 Feb 2022
Code	https://github.com/VITA-Group/AsViT
Authors	Wuyang Chen, Wei Huang, Xianzhi Du, Xiaodan Song, Zhangyang Wang, Denny Zhou
Org.	University of Texas, University of Technology Sydney, Google
Abstract	This work targets automated designing and scaling of Vision Transformers (ViTs). The motivation comes from two pain spots: 1) the lack of efficient and principled methods for designing and scaling ViTs; 2) the tremendous computational cost of training ViT that is much heavier than its convolution counterpart. To tackle these issues, we propose As-ViT, an auto-scaling framework for ViTs without training, which automatically discovers and scales up ViTs in an efficient and principled manner. Specifically, we first design a "seed" ViT topology by leveraging a training-free search process. This extremely fast search is fulfilled by a comprehensive study of ViT's network complexity, yielding a strong Kendall-tau correlation with ground-truth accuracies. Second, starting from the "seed" topology, we automate the scaling rule for ViTs by growing widths/depths to different ViT layers. This results in a series of architectures with different numbers of parameters in a single run. Finally, based on the observation that ViTs can tolerate coarse tokenization in early training stages, we propose a progressive tokenization strategy to train ViTs faster and cheaper. As a unified framework, As-ViT achieves strong performance on classification (83.5% top1 on ImageNet-1k) and detection (52.7% mAP on COCO) without any manual crafting nor scaling of ViT architectures: the end-to-end model design and scaling process cost only 12 hours on one V100 GPU.

ViT-P

Paper	ViT-P: Rethinking Data-efficient Vision Transformers from Locality
Date	4 Mar 2022
Code	https://github.com/freder-chen/vitp
Authors	Bin Chen, Ran Wang, Di Ming, Xin Feng
Org.	Chongqing University of Technology, University of California
Abstract	Recent advances of Transformers have brought new trust to computer vision tasks. However, on small dataset, Transformers is hard to train and has lower performance than convolutional neural networks. We make vision transformers as data-efficient as convolutional neural networks by introducing multi-focal attention bias. Inspired by the attention distance in a well-trained ViT, we constrain the self-attention of ViT to have multi-scale localized receptive field. The size of receptive field is adaptable during training so that optimal configuration can be learned. We provide empirical evidence that proper constrain of receptive field can reduce the amount of training data for vision transformers. On Cifar100, our ViT-P Base model achieves the state-of-the-art accuracy (83.16%) trained from scratch. We also perform analysis on ImageNet to show our method does not lose accuracy on large data sets.

CF-ViT(Coarse-to-Fine Vision Transformer)

Paper	Coarse-to-Fine Vision Transformer
Date	8 Mar 2022
Code	https://github.com/chenmnz/cf-vit
Authors	Mengzhao Chen, Mingbao Lin, Ke Li, Yunhang Shen, Yongjian Wu, Fei Chao, Rongrong Ji
Org.	Xiamen University, Tencent Youtu Lab
Abstract	Vision Transformers (ViT) have made many breakthroughs in computer vision tasks. However, considerable redundancy arises in the spatial dimension of an input image, leading to massive computational costs. Therefore, We propose a coarse-to-fine vision transformer (CF-ViT) to relieve computational burden while retaining performance in this paper. Our proposed CF-ViT is motivated by two important observations in modern ViT models: (1) The coarse-grained patch splitting can locate informative regions of an input image. (2) Most images can be well recognized by a ViT model in a small-length token sequence. Therefore, our CF-ViT implements network inference in a two-stage manner. At coarse inference stage, an input image is split into a small-length patch sequence for a computationally economical classification. If not well recognized, the informative patches are identified and further re-split in a fine-grained granularity. Extensive experiments demonstrate the efficacy of our CF-ViT. For example, without any compromise on performance, CF-ViT reduces 53% FLOPs of LV-ViT, and also achieves 2.01x throughput.

DGT(Dynamic Group Transformer)

Paper	Dynamic Group Transformer: A General Vision Transformer Backbone with Dynamic Group Attention
Date	8 Mar 2022
Code
Authors	Kai Liu, Tianyi Wu, Cong Liu, Guodong Guo
Org.	Sun Yat-sen University, Baidu Research
Abstract	Recently, Transformers have shown promising performance in various vision tasks. To reduce the quadratic computation complexity caused by each query attending to all keys/values, various methods have constrained the range of attention within local regions, where each query only attends to keys/values within a hand-crafted window. However, these hand-crafted window partition mechanisms are data-agnostic and ignore their input content, so it is likely that one query maybe attends to irrelevant keys/values. To address this issue, we propose a Dynamic Group Attention (DG-Attention), which dynamically divides all queries into multiple groups and selects the most relevant keys/values for each group. Our DG-Attention can flexibly model more relevant dependencies without any spatial constraint that is used in hand-crafted window based attention. Built on the DG-Attention, we develop a general vision transformer backbone named Dynamic Group Transformer (DGT). Extensive experiments show that our models can outperform the state-of-the-art methods on multiple common vision tasks, including image classification, semantic segmentation, object detection, and instance segmentation.

WinfT(Window-Free Transformer)

Paper	Visualizing and Understanding Patch Interactions in Vision Transformer
Date	11 Mar 2022
Code
Authors	Jie Ma, Yalong Bai, Bineng Zhong, Wei Zhang, Ting Yao, Tao Mei
Org.	Guangxi Normal University, Huaqiao University, JD Explore Academy
Abstract	Vision Transformer (ViT) has become a leading tool in various computer vision tasks, owing to its unique self-attention mechanism that learns visual representations explicitly through cross-patch information interactions. Despite having good success, the literature seldom explores the explainability of vision transformer, and there is no clear picture of how the attention mechanism with respect to the correlation across comprehensive patches will impact the performance and what is the further potential. In this work, we propose a novel explainable visualization approach to analyze and interpret the crucial attention interactions among patches for vision transformer. Specifically, we first introduce a quantification indicator to measure the impact of patch interaction and verify such quantification on attention window design and indiscriminative patches removal. Then, we exploit the effective responsive field of each patch in ViT and devise a window-free transformer architecture accordingly. Extensive experiments on ImageNet demonstrate that the exquisitely designed quantitative method is shown able to facilitate ViT model learning, leading the top-1 accuracy by 4.28% at most. Moreover, the results on downstream fine-grained recognition tasks further validate the generalization of our proposal.

EIT(Efficiently lead Inductive biases to ViT)

Paper	EIT: Efficiently Lead Inductive Biases to ViT
Date	14 Mar 2022
Code	https://github.com/MrHaiPi/EIT
Authors	Rui Xia, Jingchao Wang, Chao Xue, Boyu Deng, Fang Wang
Org.	Academy of Military Sciences, Peng Cheng Laboratory, Tianjin Artificial Intelligence Innovation Center
Abstract	Vision Transformer (ViT) depends on properties similar to the inductive bias inherent in Convolutional Neural Networks to perform better on non-ultra-large scale datasets. In this paper, we propose an architecture called Efficiently lead Inductive biases to ViT (EIT), which can effectively lead the inductive biases to both phases of ViT. In the Patches Projection phase, a convolutional max-pooling structure is used to produce overlapping patches. In the Transformer Encoder phase, we design a novel inductive bias introduction structure called decreasing convolution, which is introduced parallel to the multi-headed attention module, by which the embedding's different channels are processed respectively. In four popular small-scale datasets, compared with ViT, EIT has an accuracy improvement of 12.6% on average with fewer parameters and FLOPs. Compared with ResNet, EIT exhibits higher accuracy with only 17.7% parameters and fewer FLOPs. Finally, ablation studies show that the EIT is efficient and does not require position embedding.

ScalableViT

Paper	ScalableViT: Rethinking the Context-oriented Generalization of Vision Transformer
Date	21 Mar 2022
Code
Authors	Rui Yang, Hailong Ma, Jie Wu, Yansong Tang, Xuefeng Xiao, Min Zheng, Xiu Li
Org.	Tsinghua University, ByteDance Inc.
Abstract	The vanilla self-attention mechanism inherently relies on pre-defined and steadfast computational dimensions. Such inflexibility restricts it from possessing context-oriented generalization that can bring more contextual cues and global representations. To mitigate this issue, we propose a Scalable Self-Attention (SSA) mechanism that leverages two scaling factors to release dimensions of query, key, and value matrix while unbinding them with the input. This scalability fetches context-oriented generalization and enhances object sensitivity, which pushes the whole network into a more effective trade-off state between accuracy and cost. Furthermore, we propose an Interactive Window-based Self-Attention (IWSA), which establishes interaction between non-overlapping regions by re-merging independent value tokens and aggregating spatial information from adjacent windows. By stacking the SSA and IWSA alternately, the Scalable Vision Transformer (ScalableViT) achieves state-of-the-art performance in general-purpose vision tasks. For example, ScalableViT-S outperforms Twins-SVT-S by 1.4% and Swin-T by 1.8% on ImageNet-1K classification.

FocalNet

Paper	Focal Modulation Networks
Date	22 Mar 2022
Code	https://github.com/microsoft/FocalNet
Authors	Jianwei Yang, Chunyuan Li, Jianfeng Gao
Org.	Microsoft Research
Abstract	In this work, we propose focal modulation network (FocalNet in short), where self-attention (SA) is completely replaced by a focal modulation module that is more effective and efficient for modeling token interactions. Focal modulation comprises three components: (i) hierarchical contextualization, implemented using a stack of depth-wise convolutional layers, to encode visual contexts from short to long ranges at different granularity levels, (ii) gated aggregation to selectively aggregate context features for each visual token (query) based on its content, and (iii) modulation or element-wise affine transformation to fuse the aggregated features into the query vector. Extensive experiments show that FocalNets outperform the state-of-the-art SA counterparts (e.g., Swin Transformers) with similar time and memory cost on the tasks of image classification, object detection, and semantic segmentation. Specifically, our FocalNets with tiny and base sizes achieve 82.3% and 83.9% top-1 accuracy on ImageNet-1K. After pretrained on ImageNet-22K, it attains 86.5% and 87.3% top-1 accuracy when finetuned with resolution 224×224 and 384×384, respectively. FocalNets exhibit remarkable superiority when transferred to downstream tasks. For object detection with Mask R-CNN, our FocalNet base trained with 1× already surpasses Swin trained with 3× schedule (49.0 v.s. 48.5). For semantic segmentation with UperNet, FocalNet base evaluated at single-scale outperforms Swin evaluated at multi-scale (50.5 v.s. 49.7). These results render focal modulation a favorable alternative to SA for effective and efficient visual modeling in real-world applications.

DW-ViT(Dynamic Window Vision Transformer)

Paper	Beyond Fixation: Dynamic Window Visual Transformer
Date	24 Mar 2022
Code	https://github.com/pzhren/dw-vit
Authors	Pengzhen Ren, Changlin Li, Guangrun Wang, Yun Xiao, Qing Du Xiaodan Liang Xiaojun Chang
Org.	Northwest University, University of Technology Sydney, University of Oxford, South China University of Technology, Sun Yat-sen University, PengCheng Laboratory, RMIT University
Abstract	Recently, a surge of interest in visual transformers is to reduce the computational cost by limiting the calculation of self-attention to a local window. Most current work uses a fixed single-scale window for modeling by default, ignoring the impact of window size on model performance. However, this may limit the modeling potential of these window-based models for multi-scale information. In this paper, we propose a novel method, named Dynamic Window Vision Transformer (DW-ViT). The dynamic window strategy proposed by DW-ViT goes beyond the model that employs a fixed single window setting. To the best of our knowledge, we are the first to use dynamic multi-scale windows to explore the upper limit of the effect of window settings on model performance. In DW-ViT, multi-scale information is obtained by assigning windows of different sizes to different head groups of window multi-head self-attention. Then, the information is dynamically fused by assigning different weights to the multi-scale window branches. We conducted a detailed performance evaluation on three datasets, ImageNet-1K, ADE20K, and COCO. Compared with related state-of-the-art (SoTA) methods, DW-ViT obtains the best performance. Specifically, compared with the current SoTA Swin Transformers \cite{liu2021swin}, DW-ViT has achieved consistent and substantial improvements on all three datasets with similar parameters and computational costs. In addition, DW-ViT exhibits good scalability and can be easily inserted into any window-based visual transformers.

SepViT(Separable Vision Transformer)

Paper	SepViT: Separable Vision Transformer
Date	29 Mar 2022
Code
Authors	Wei Li, Xing Wang, Xin Xia, Jie Wu, Xuefeng Xiao, Min Zheng, Shiping Wen
Org.	University of Electronic Science and Technology of China, ByteDance Inc., University of Technology Sydney
Abstract	Vision Transformers have witnessed prevailing success in a series of vision tasks. However, they often require enormous amount of computations to achieve high performance, which is burdensome to deploy on resource-constrained devices. To address these issues, we draw lessons from depthwise separable convolution and imitate its ideology to design the Separable Vision Transformer, abbreviated as SepViT. SepViT helps to carry out the information interaction within and among the windows via a depthwise separable self-attention. The novel window token embedding and grouped self-attention are employed to model the attention relationship among windows with negligible computational cost and capture a long-range visual dependencies of multiple windows, respectively. Extensive experiments on various benchmark tasks demonstrate SepViT can achieve state-of-the-art results in terms of trade-off between accuracy and latency. Among them, SepViT achieves 84.0% top-1 accuracy on ImageNet-1K classification while decreasing the latency by 40%, compared to the ones with similar accuracy (e.g., CSWin, PVTV2). As for the downstream vision tasks, SepViT with fewer FLOPs can achieve 50.4% mIoU on ADE20K semantic segmentation task, 47.5 AP on the RetinaNet-based COCO detection task, 48.7 box AP and 43.9 mask AP on Mask R-CNN-based COCO detection and segmentation tasks.

2022 Q2

MaxViT(Multi-Axis ViT)

Paper	MaxViT: Multi-Axis Vision Transformer
Date	4 Apr 2022
Code
Authors	Zhengzhong Tu, Hossein Talebi, Han Zhang, Feng Yang, Peyman Milanfar, Alan Bovik, Yinxiao Li
Org.	Google Research, University of Texas at Austin
Abstract	Transformers have recently gained significant attention in the computer vision community. However, the lack of scalability of self-attention mechanisms with respect to image size has limited their wide adoption in state-of-the-art vision backbones. In this paper we introduce an efficient and scalable attention model we call multi-axis attention, which consists of two aspects: blocked local and dilated global attention. These design choices allow global-local spatial interactions on arbitrary input resolutions with only linear complexity. We also present a new architectural element by effectively blending our proposed attention model with convolutions, and accordingly propose a simple hierarchical vision backbone, dubbed MaxViT, by simply repeating the basic building block over multiple stages. Notably, MaxViT is able to "see" globally throughout the entire network, even in earlier, high-resolution stages. We demonstrate the effectiveness of our model on a broad spectrum of vision tasks. On image classification, MaxViT achieves state-of-the-art performance under various settings: without extra data, MaxViT attains 86.5% ImageNet-1K top-1 accuracy; with ImageNet-21K pre-training, our model achieves 88.7% top-1 accuracy. For downstream tasks, MaxViT as a backbone delivers favorable performance on object detection as well as visual aesthetic assessment. We also show that our proposed model expresses strong generative modeling capability on ImageNet, demonstrating the superior potential of MaxViT blocks as a universal vision module.

DaVit(Dual Attention Vision Transformers)

Paper	DaViT: Dual Attention Vision Transformers
Date	7 Apr 2022
Code	https://arxiv.org/abs/2204.03645v1
Authors	Mingyu Ding, Bin Xiao, Noel Codella, Ping Luo, Jingdong Wang, Lu Yuan
Org.	The University of Hong Kong, Microsoft Cloud + AI , Baidu
Abstract	In this work, we introduce Dual Attention Vision Transformers (DaViT), a simple yet effective vision transformer architecture that is able to capture global context while maintaining computational efficiency. We propose approaching the problem from an orthogonal angle: exploiting self-attention mechanisms with both "spatial tokens" and "channel tokens". With spatial tokens, the spatial dimension defines the token scope, and the channel dimension defines the token feature dimension. With channel tokens, we have the inverse: the channel dimension defines the token scope, and the spatial dimension defines the token feature dimension. We further group tokens along the sequence direction for both spatial and channel tokens to maintain the linear complexity of the entire model. We show that these two self-attentions complement each other: (i) since each channel token contains an abstract representation of the entire image, the channel attention naturally captures global interactions and representations by taking all spatial positions into account when computing attention scores between channels; (ii) the spatial attention refines the local representations by performing fine-grained interactions across spatial locations, which in turn helps the global information modeling in channel attention. Extensive experiments show our DaViT achieves state-of-the-art performance on four different tasks with efficient computations. Without extra data, DaViT-Tiny, DaViT-Small, and DaViT-Base achieve 82.8%, 84.2%, and 84.6% top-1 accuracy on ImageNet-1K with 28.3M, 49.7M, and 87.9M parameters, respectively. When we further scale up DaViT with 1.5B weakly supervised image and text pairs, DaViT-Gaint reaches 90.4% top-1 accuracy on ImageNet-1K.

DeiT III

Paper	DeiT III: Revenge of the ViT
Date	14 Apr 2022
Code
Authors	Hugo Touvron, Matthieu Cord, Hervé Jégou
Org.	Meta AI, Sorbonne University
Abstract	A Vision Transformer (ViT) is a simple neural architecture amenable to serve several computer vision tasks. It has limited built-in architectural priors, in contrast to more recent architectures that incorporate priors either about the input data or of specific tasks. Recent works show that ViTs benefit from self-supervised pre-training, in particular BerT-like pre-training like BeiT. In this paper, we revisit the supervised training of ViTs. Our procedure builds upon and simplifies a recipe introduced for training ResNet-50. It includes a new simple data-augmentation procedure with only 3 augmentations, closer to the practice in self-supervised learning. Our evaluations on Image classification (ImageNet-1k with and without pre-training on ImageNet-21k), transfer learning and semantic segmentation show that our procedure outperforms by a large margin previous fully supervised training recipes for ViT. It also reveals that the performance of our ViT trained with supervision is comparable to that of more recent architectures. Our results could serve as better baselines for recent self-supervised approaches demonstrated on ViT.

NAT(Neighborhood Attention Transformer)

Paper	Neighborhood Attention Transformer
Date	14 Apr 2022
Code	https://github.com/SHI-Labs/Neighborhood-Attention-Transformer
Authors	Ali Hassani, Steven Walton, Jiachen Li, Shen Li, Humphrey Shi
Org.	SHI Lab @ U of Oregon & UIUC, Picsart AI Research,Meta/Facebook AI
Abstract	We present Neighborhood Attention Transformer (NAT), an efficient, accurate and scalable hierarchical transformer that works well on both image classification and downstream vision tasks. It is built upon Neighborhood Attention (NA), a simple and flexible attention mechanism that localizes the receptive field for each query to its nearest neighboring pixels. NA is a localization of self-attention, and approaches it as the receptive field size increases. It is also equivalent in FLOPs and memory usage to Swin Transformer's shifted window attention given the same receptive field size, while being less constrained. Furthermore, NA includes local inductive biases, which eliminate the need for extra operations such as pixel shifts. Experimental results on NAT are competitive; NAT-Tiny reaches 83.2% top-1 accuracy on ImageNet with only 4.3 GFLOPs and 28M parameters, 51.4% mAP on MS-COCO and 48.4% mIoU on ADE20k.

Rest V2(ResNet Transformer V2)

Paper	ResT V2: Simpler, Faster and Stronger
Date	15 Apr 2022
Code	https://github.com/wofmanaf/ResT
Authors	Qing-Long Zhang, Yu-Bin Yang
Org.	State Key Laboratory for Novel Software Technology Nanjing University
Abstract	This paper proposes ResTv2, a simpler, faster, and stronger multi-scale vision Transformer for visual recognition. ResTv2 simplifies the EMSA structure in ResTv1 (i.e., eliminating the multi-head interaction part) and employs an upsample operation to reconstruct the lost medium- and high-frequency information caused by the downsampling operation. In addition, we explore different techniques for better apply ResTv2 backbones to downstream tasks. We found that although combining EMSAv2 and window attention can greatly reduce the theoretical matrix multiply FLOPs, it may significantly decrease the computation density, thus causing lower actual speed. We comprehensively validate ResTv2 on ImageNet classification, COCO detection, and ADE20K semantic segmentation. Experimental results show that the proposed ResTv2 can outperform the recently state-of-the-art backbones by a large margin, demonstrating the potential of ResTv2 as solid backbones.

ASF-former

Paper	Adaptive Split-Fusion Transformer
Date	26 Apr 2022
Code	https://github.com/szx503045266/ASF-former
Authors	Zixuan Su, Hao Zhang, Jingjing Chen, Lei Pang, Chong-Wah Ngo, Yu-Gang Jiang
Org.	Fudan University, Singapore Management University,City University of Hong Kong
Abstract	Neural networks for visual content understanding have recently evolved from convolutional ones (CNNs) to transformers. The prior (CNN) relies on small-windowed kernels to capture the regional clues, demonstrating solid local expressiveness. On the contrary, the latter (transformer) establishes long-range global connections between localities for holistic learning. Inspired by this complementary nature, there is a growing interest in designing hybrid models to best utilize each technique. Current hybrids merely replace convolutions as simple approximations of linear projection or juxtapose a convolution branch with attention, without concerning the importance of local/global modeling. To tackle this, we propose a new hybrid named Adaptive Split-Fusion Transformer (ASF-former) to treat convolutional and attention branches differently with adaptive weights. Specifically, an ASF-former encoder equally splits feature channels into half to fit dual-path inputs. Then, the outputs of dual-path are fused with weighting scalars calculated from visual cues. We also design the convolutional path compactly for efficiency concerns. Extensive experiments on standard benchmarks, such as ImageNet-1K, CIFAR-10, and CIFAR-100, show that our ASF-former outperforms its CNN, transformer counterparts, and hybrid pilots in terms of accuracy (83.9% on ImageNet-1K), under similar conditions (12.9G MACs/56.7M Params, without large-scale pre-training).

EdgeViTs

Paper	EdgeViTs: Competing Light-weight CNNs on Mobile Devices with Vision Transformers
Date	6 May 2022
Code
Authors	Junting Pan, Adrian Bulat, Fuwen Tan, Xiatian Zhu, Lukasz Dudziak, Hongsheng Li, Georgios Tzimiropoulos, Brais Martinez
Org.	The Chinese University of Hong Kong, Samsung AI Cambridge, Queen Mary University of London
Abstract	Self-attention based models such as vision transformers (ViTs) have emerged as a very competitive architecture alternative to convolutional neural networks (CNNs) in computer vision. Despite increasingly stronger variants with ever-higher recognition accuracies, due to the quadratic complexity of self-attention, existing ViTs are typically demanding in computation and model size. Although several successful design choices (e.g., the convolutions and hierarchical multi-stage structure) of prior CNNs have been reintroduced into recent ViTs, they are still not sufficient to meet the limited resource requirements of mobile devices. This motivates a very recent attempt to develop light ViTs based on the state-of-the-art MobileNet-v2, but still leaves a performance gap behind. In this work, pushing further along this under-studied direction we introduce EdgeViTs, a new family of light-weight ViTs that, for the first time, enable attention-based vision models to compete with the best light-weight CNNs in the tradeoff between accuracy and on-device efficiency. This is realized by introducing a highly cost-effective local-global-local (LGL) information exchange bottleneck based on optimal integration of self-attention and convolutions. For device-dedicated evaluation, rather than relying on inaccurate proxies like the number of FLOPs or parameters, we adopt a practical approach of focusing directly on on-device latency and, for the first time, energy efficiency. Specifically, we show that our models are Pareto-optimal when both accuracy-latency and accuracy-energy trade-offs are considered, achieving strict dominance over other ViTs in almost all cases and competing with the most efficient CNNs.

TRT-Vit

Paper	TRT-ViT: TensorRT-oriented Vision Transformer
Date	19 May 2022
Code
Authors	Xin Xia, Jiashi Li, Jie Wu, Xing Wang, Xuefeng Xiao, Min Zheng, Rui Wang
Org.	ByteDance Inc.
Abstract	We revisit the existing excellent Transformers from the perspective of practical application. Most of them are not even as efficient as the basic ResNets series and deviate from the realistic deployment scenario. It may be due to the current criterion to measure computation efficiency, such as FLOPs or parameters is one-sided, sub-optimal, and hardware-insensitive. Thus, this paper directly treats the TensorRT latency on the specific hardware as an efficiency metric, which provides more comprehensive feedback involving computational capacity, memory cost, and bandwidth. Based on a series of controlled experiments, this work derives four practical guidelines for TensorRT-oriented and deployment-friendly network design, e.g., early CNN and late Transformer at stage-level, early Transformer and late CNN at block-level. Accordingly, a family of TensortRT-oriented Transformers is presented, abbreviated as TRT-ViT. Extensive experiments demonstrate that TRT-ViT significantly outperforms existing ConvNets and vision Transformers with respect to the latency/accuracy trade-off across diverse visual tasks, e.g., image classification, object detection and semantic segmentation. For example, at 82.7% ImageNet-1k top-1 accuracy, TRT-ViT is 2.7× faster than CSWin and 2.0× faster than Twins. On the MS-COCO object detection task, TRT-ViT achieves comparable performance with Twins, while the inference speed is increased by 2.8×.

SuperVit

Paper	Super Vision Transformer
Date	23 May 2022
Code	https://github.com/lmbxmu/SuperViT
Authors	Mingbao Lin, Mengzhao Chen, Yuxin Zhang, Ke Li, Yunhang Shen, Chunhua Shen, Rongrong Ji
Org.	Xiamen,Tencent Youtu Lab, Zhejiang University
Abstract	We attempt to reduce the computational costs in vision transformers (ViTs), which increase quadratically in the token number. We present a novel training paradigm that trains only one ViT model at a time, but is capable of providing improved image recognition performance with various computational costs. Here, the trained ViT model, termed super vision transformer (SuperViT), is empowered with the versatile ability to solve incoming patches of multiple sizes as well as preserve informative tokens with multiple keeping rates (the ratio of keeping tokens) to achieve good hardware efficiency for inference, given that the available hardware resources often change from time to time. Experimental results on ImageNet demonstrate that our SuperViT can considerably reduce the computational costs of ViT models with even performance increase. For example, we reduce 2x FLOPs of DeiT-S while increasing the Top-1 accuracy by 0.2% and 0.7% for 1.5x reduction. Also, our SuperViT significantly outperforms existing studies on efficient vision transformers. For example, when consuming the same amount of FLOPs, our SuperViT surpasses the recent state-of-the-art (SoTA) EViT by 1.1% when using DeiT-S as their backbones.

iFormer(Inception Transformer)

Paper	Inception Transformer
Date	25 May 2022
Code	https://github.com/sail-sg/iFormer
Authors	Chenyang Si, Weihao Yu, Pan Zhou, Yichen Zhou, Xinchao Wang, Shuicheng Yan
Org.	Sea AI Lab 2National University of Singapore
Abstract	Recent studies show that Transformer has strong capability of building long-range dependencies, yet is incompetent in capturing high frequencies that predominantly convey local information. To tackle this issue, we present a novel and general-purpose Inception Transformer, or iFormer for short, that effectively learns comprehensive features with both high- and low-frequency information in visual data. Specifically, we design an Inception mixer to explicitly graft the advantages of convolution and max-pooling for capturing the high-frequency information to Transformers. Different from recent hybrid frameworks, the Inception mixer brings greater efficiency through a channel splitting mechanism to adopt parallel convolution/max-pooling path and self-attention path as high- and low-frequency mixers, while having the flexibility to model discriminative information scattered within a wide frequency range. Considering that bottom layers play more roles in capturing high-frequency details while top layers more in modeling low-frequency global information, we further introduce a frequency ramp structure, i.e. gradually decreasing the dimensions fed to the high-frequency mixer and increasing those to the low-frequency mixer, which can effectively trade-off high- and low-frequency components across different layers. We benchmark the iFormer on a series of vision tasks, and showcase that it achieves impressive performance on image classification, COCO detection and ADE20K segmentation. For example, our iFormer-S hits the top-1 accuracy of 83.4% on ImageNet-1K, much higher than DeiT-S by 3.6%, and even slightly better than much bigger model Swin-B (83.3%) with only 1/4 parameters and 1/3 FLOPs.

LITv2(Less attention vIsion Transformer V2)

Paper	Fast Vision Transformers with HiLo Attention
Date	26 May 2022
Code	https://github.com/zip-group/LITv2
Authors	Zizheng Pan, Jianfei Cai, Bohan Zhuang
Org.	Data Science & AI, Monash University
Abstract	Vision Transformers (ViTs) have triggered the most recent and significant breakthroughs in computer vision. Their efficient designs are mostly guided by the indirect metric of computational complexity, i.e., FLOPs, which however has a clear gap with the direct metric such as throughput. Thus, we propose to use the direct speed evaluation on the target platform as the design principle for efficient ViTs. Particularly, we introduce LITv2, a simple and effective ViT which performs favourably against the existing state-of-the-art methods across a spectrum of different model sizes with faster speed. At the core of LITv2 is a novel self-attention mechanism, which we dub HiLo. HiLo is inspired by the insight that high frequencies in an image capture local fine details and low frequencies focus on global structures, whereas a multi-head self-attention layer neglects the characteristic of different frequencies. Therefore, we propose to disentangle the high/low frequency patterns in an attention layer by separating the heads into two groups, where one group encodes high frequencies via self-attention within each local window, and another group performs the attention to model the global relationship between the average-pooled low-frequency keys from each window and each query position in the input feature map. Benefit from the efficient design for both groups, we show that HiLo is superior to the existing attention mechanisms by comprehensively benchmarking on FLOPs, speed and memory consumption on GPUs. Powered by HiLo, LITv2 serves as a strong backbone for mainstream vision tasks including image classification, dense detection and segmentation.

AdaptFormer

Paper	AdaptFormer: Adapting Vision Transformers for Scalable Visual Recognition
Date	26 May 2022
Code	http://www.shoufachen.com/adaptformer-page
Authors	Shoufa Chen, Chongjian Ge, Zhan Tong, Jiangliu Wang, Yibing Song, Jue Wang, Ping Luo
Org.	The University of Hong Kong / Tencent AI Lab / The Chinese University of Hong Kong
Abstract	Although the pre-trained Vision Transformers (ViTs) achieved great success in computer vision, adapting a ViT to various image and video tasks is challenging because of its heavy computation and storage burdens, where each model needs to be independently and comprehensively fine-tuned to different tasks, limiting its transferability in different domains. To address this challenge, we propose an effective adaptation approach for Transformer, namely AdaptFormer, which can adapt the pre-trained ViTs into many different image and video tasks efficiently. It possesses several benefits more appealing than prior arts. Firstly, AdaptFormer introduces lightweight modules that only add less than 2% extra parameters to a ViT, while it is able to increase the ViT's transferability without updating its original pre-trained parameters, significantly outperforming the existing 100% fully fine-tuned models on action recognition benchmarks. Secondly, it can be plug-and-play in different Transformers and scalable to many visual tasks. Thirdly, extensive experiments on five image and video datasets show that AdaptFormer largely improves ViTs in the target domains. For example, when updating just 1.5% extra parameters, it achieves about 10% and 19% relative improvement compared to the fully fine-tuned models on Something-Something~v2 and HMDB51, respectively.

X-ViT

Paper	X-ViT: High Performance Linear Vision Transformer without Softmax
Date	27 May 2022
Code
Authors	Jeonggeun Song, Heung-Chang Lee
Org.	Kakao Enterprise
Abstract	Vision transformers have become one of the most important models for computer vision tasks. Although they outperform prior works, they require heavy computational resources on a scale that is quadratic to the number of tokens, N. This is a major drawback of the traditional self-attention (SA) algorithm. Here, we propose the X-ViT, ViT with a novel SA mechanism that has linear complexity. The main approach of this work is to eliminate nonlinearity from the original SA. We factorize the matrix multiplication of the SA mechanism without complicated linear approximation. By modifying only a few lines of code from the original SA, the proposed models outperform most transformer-based models on image classification and dense prediction tasks on most capacity regimes.

HiViT(Hierarchical Vision Transformer)

Paper	HiViT: Hierarchical Vision Transformer Meets Masked Image Modeling
Date	30 May 2022
Code
Authors	Xiaosong Zhang, Yunjie Tian, Wei Huang, Qixiang Ye, Qi Dai, Lingxi Xie, Qi Tian
Org.	University of Chinese Academy of Sciences / Fudan University / Huawei Inc
Abstract	Recently, masked image modeling (MIM) has offered a new methodology of self-supervised pre-training of vision transformers. A key idea of efficient implementation is to discard the masked image patches (or tokens) throughout the target network (encoder), which requires the encoder to be a plain vision transformer (e.g., ViT), albeit hierarchical vision transformers (e.g., Swin Transformer) have potentially better properties in formulating vision inputs. In this paper, we offer a new design of hierarchical vision transformers named HiViT (short for Hierarchical ViT) that enjoys both high efficiency and good performance in MIM. The key is to remove the unnecessary "local inter-unit operations", deriving structurally simple hierarchical vision transformers in which mask-units can be serialized like plain vision transformers. For this purpose, we start with Swin Transformer and (i) set the masking unit size to be the token size in the main stage of Swin Transformer, (ii) switch off inter-unit self-attentions before the main stage, and (iii) eliminate all operations after the main stage. Empirical studies demonstrate the advantageous performance of HiViT in terms of fully-supervised, self-supervised, and transfer learning. In particular, in running MAE on ImageNet-1K, HiViT-B reports a +0.6% accuracy gain over ViT-B and a 1.9× speed-up over Swin-B, and the performance gain generalizes to downstream tasks of detection and segmentation. Code will be made publicly available.

EfficientViT

Paper	EfficientViT: Enhanced Linear Attention for High-Resolution Low-Computation Visual Recognition
Date	29 May 2022
Code
Authors	Han Cai, Chuang Gan, Song Han
Org.	MIT , MIT-IBM Watson AI Lab
Abstract	Vision Transformer (ViT) has achieved remarkable performance in many vision tasks. However, ViT is inferior to convolutional neural networks (CNNs) when targeting high-resolution mobile vision applications. The key computational bottleneck of ViT is the softmax attention module which has quadratic computational complexity with the input resolution. It is essential to reduce the cost of ViT to deploy it on edge devices. Existing methods (e.g., Swin, PVT) restrict the softmax attention within local windows or reduce the resolution of key/value tensors to reduce the cost, which sacrifices ViT's core advantages on global feature extractions. In this work, we present EfficientViT, an efficient ViT architecture for high-resolution low-computation visual recognition. Instead of restricting the softmax attention, we propose to replace softmax attention with linear attention while enhancing its local feature extraction ability with depthwise convolution. EfficientViT maintains global and local feature extraction capability while enjoying linear computational complexity. Extensive experiments on COCO object detection and Cityscapes semantic segmentation demonstrate the effectiveness of our method. On the COCO dataset, EfficientViT achieves 42.6 AP with 4.4G MACs, surpassing EfficientDet-D1 by 2.4 AP while having 27.9% fewer MACs. On Cityscapes, EfficientViT reaches 78.7 mIoU with 19.1G MACs, outperforming SegFormer by 2.5 mIoU while requiring less than 1/3 the computational cost. On Qualcomm Snapdragon 855 CPU, EfficientViT is 3x faster than EfficientNet while achieving higher ImageNet accuracy.

EfficientFormer

Paper	EfficientFormer: Vision Transformers at MobileNet Speed
Date	2 Jun 2022
Code	https://github.com/snap-research/efficientformer
Authors	Yanyu Li, Geng Yuan, Yang Wen, Eric Hu, Georgios Evangelidis, Sergey Tulyakov, Yanzhi Wang, Jian Ren
Org.	Snap Inc. , Northeastern University
Abstract	Vision Transformers (ViT) have shown rapid progress in computer vision tasks, achieving promising results on various benchmarks. However, due to the massive number of parameters and model design, e.g., attention mechanism, ViT-based models are generally times slower than lightweight convolutional networks. Therefore, the deployment of ViT for real-time applications is particularly challenging, especially on resource-constrained hardware such as mobile devices. Recent efforts try to reduce the computation complexity of ViT through network architecture search or hybrid design with MobileNet block, yet the inference speed is still unsatisfactory. This leads to an important question: can transformers run as fast as MobileNet while obtaining high performance? To answer this, we first revisit the network architecture and operators used in ViT-based models and identify inefficient designs. Then we introduce a dimension-consistent pure transformer (without MobileNet blocks) as design paradigm. Finally, we perform latency-driven slimming to get a series of final models dubbed EfficientFormer. Extensive experiments show the superiority of EfficientFormer in performance and speed on mobile devices. Our fastest model, EfficientFormer-L1, achieves 79.2% top-1 accuracy on ImageNet-1K with only 1.6 ms inference latency on iPhone 12 (compiled with CoreML), which is even a bit faster than MobileNetV2 (1.7 ms, 71.8% top-1), and our largest model, EfficientFormer-L7, obtains 83.3% accuracy with only 7.0 ms latency. Our work proves that properly designed transformers can reach extremely low latency on mobile devices while maintaining high performance

MobileViTv2

Paper	Separable Self-attention for Mobile Vision Transformers
Date	6 Jun 2022
Code	https://github.com/apple/ml-cvnets
Authors	Sachin Mehta, Mohammad Rastegari
Org.	Apple
Abstract	Mobile vision transformers (MobileViT) can achieve state-of-the-art performance across several mobile vision tasks, including classification and detection. Though these models have fewer parameters, they have high latency as compared to convolutional neural network-based models. The main efficiency bottleneck in MobileViT is the multi-headed self-attention (MHA) in transformers, which requires O(k2) time complexity with respect to the number of tokens (or patches) k. Moreover, MHA requires costly operations (e.g., batch-wise matrix multiplication) for computing self-attention, impacting latency on resource-constrained devices. This paper introduces a separable self-attention method with linear complexity, i.e. O(k). A simple yet effective characteristic of the proposed method is that it uses element-wise operations for computing self-attention, making it a good choice for resource-constrained devices. The improved model, MobileViTv2, is state-of-the-art on several mobile vision tasks, including ImageNet object classification and MS-COCO object detection. With about three million parameters, MobileViTv2 achieves a top-1 accuracy of 75.6% on the ImageNet dataset, outperforming MobileViT by about 1% while running 3.2× faster on a mobile device.

PerViT(Peripheral Vision Transformer)

Paper	Peripheral Vision Transformer
Date	14 Jun 2022
Code
Authors	Juhong Min, Yucheng Zhao, Chong Luo, Minsu Cho
Org.	Pohang University of Science and Technology, Microsoft Research Asia , University of Science and Technology of China
Abstract	Human vision possesses a special type of visual processing systems called peripheral vision. Partitioning the entire visual field into multiple contour regions based on the distance to the center of our gaze, the peripheral vision provides us the ability to perceive various visual features at different regions. In this work, we take a biologically inspired approach and explore to model peripheral vision in deep neural networks for visual recognition. We propose to incorporate peripheral position encoding to the multi-head self-attention layers to let the network learn to partition the visual field into diverse peripheral regions given training data. We evaluate the proposed network, dubbed PerViT, on the large-scale ImageNet dataset and systematically investigate the inner workings of the model for machine perception, showing that the network learns to perceive visual data similarly to the way that human vision does. The state-of-the-art performance in image classification task across various model sizes demonstrates the efficacy of the proposed method.

SP-ViT(Spatial Priors ViT)

Paper	SP-ViT: Learning 2D Spatial Priors for Vision Transformers
Date	15 Jun 2022
Code
Authors	Yuxuan Zhou, Wangmeng Xiang, Chao Li, Biao Wang, Xihan Wei, Lei Zhang, Margret Keuper, Xiansheng Hua
Org.	University of Mannheim , The Hong Kong Polytechnic University, University of Siegen, Alibaba Group
Abstract	Recently, transformers have shown great potential in image classification and established state-of-the-art results on the ImageNet benchmark. However, compared to CNNs, transformers converge slowly and are prone to overfitting in low-data regimes due to the lack of spatial inductive biases. Such spatial inductive biases can be especially beneficial since the 2D structure of an input image is not well preserved in transformers. In this work, we present Spatial Prior-enhanced Self-Attention (SP-SA), a novel variant of vanilla Self-Attention (SA) tailored for vision transformers. Spatial Priors (SPs) are our proposed family of inductive biases that highlight certain groups of spatial relations. Unlike convolutional inductive biases, which are forced to focus exclusively on hard-coded local regions, our proposed SPs are learned by the model itself and take a variety of spatial relations into account. Specifically, the attention score is calculated with emphasis on certain kinds of spatial relations at each head, and such learned spatial foci can be complementary to each other. Based on SP-SA we propose the SP-ViT family, which consistently outperforms other ViT models with similar GFlops or parameters. Our largest model SP-ViT-L achieves a record-breaking 86.3% Top-1 accuracy with a reduction in the number of parameters by almost 50% compared to previous state-of-the-art model (150M for SP-ViT-L vs 271M for CaiT-M-36) among all ImageNet-1K models trained on 224x224 and fine-tuned on 384x384 resolution w/o extra data.

EATFormer(Evolutionary Algorithm Transformer)

Paper	EATFormer: Improving Vision Transformer Inspired by Evolutionary Algorithm
Date	19 Jun 2022
Code	https://github.com/zhangzjn/eatformer
Authors	Jiangning Zhang, Xiangtai Li, Yabiao Wang, Chengjie Wang, Yibo Yang, Yong Liu, Dacheng Tao
Org.	Zhejiang Univearsity, Peking University, Youtu Lab,The University of Sydney
Abstract	Motivated by biological evolution, this paper explains the rationality of Vision Transformer by analogy with the proven practical Evolutionary Algorithm (EA) and derives that both have consistent mathematical formulation. Then inspired by effective EA variants, we propose a novel pyramid EATFormer backbone that only contains the proposed \emph{EA-based Transformer} (EAT) block, which consists of three residual parts, \ie, \emph{Multi-Scale Region Aggregation} (MSRA), \emph{Global and Local Interaction} (GLI), and \emph{Feed-Forward Network} (FFN) modules, to model multi-scale, interactive, and individual information separately. Moreover, we design a \emph{Task-Related Head} (TRH) docked with transformer backbone to complete final information fusion more flexibly and \emph{improve} a \emph{Modulated Deformable MSA} (MD-MSA) to dynamically model irregular locations. Massive quantitative and quantitative experiments on image classification, downstream tasks, and explanatory experiments demonstrate the effectiveness and superiority of our approach over State-Of-The-Art (SOTA) methods. \Eg, our Mobile (1.8M), Tiny (6.1M), Small (24.3M), and Base (49.0M) models achieve 69.4, 78.4, 83.1, and 83.9 Top-1 only trained on ImageNet-1K with naive training recipe; EATFormer-Tiny/Small/Base armed Mask-R-CNN obtain 45.4/47.4/49.0 box AP and 41.4/42.9/44.2 mask AP on COCO detection, surpassing contemporary MPViT-T, Swin-T, and Swin-S by 0.6/1.4/0.5 box AP and 0.4/1.3/0.9 mask AP separately with less FLOPs; Our EATFormer-Small/Base achieve 47.3/49.3 mIoU on ADE20K by Upernet that exceeds Swin-T/S by 2.8/1.7.

GC ViT(Global Context Vision Transformers)

Paper	Global Context Vision Transformers
Date	20 Jun 2022
Code	https://github.com/NVlabs/GCViT
Authors	Ali Hatamizadeh, Hongxu Yin, Jan Kautz, Pavlo Molchanov
Org.	NVIDIA
Abstract	We propose global context vision transformer (GC ViT), a novel architecture that enhances parameter and compute utilization. Our method leverages global context self-attention modules, joint with local self-attention, to effectively yet efficiently model both long and short-range spatial interactions, without the need for expensive operations such as computing attention masks or shifting local windows. In addition, we address the issue of lack of the inductive bias in ViTs via proposing to use a modified fused inverted residual blocks in our architecture. Our proposed GC ViT achieves state-of-the-art results across image classification, object detection and semantic segmentation tasks. On ImageNet-1K dataset for classification, the base, small and tiny variants of GC ViT with 28M, 51M and 90M parameters achieve 83.2%, 83.9% and 84.4% Top-1 accuracy, respectively, surpassing comparably-sized prior art such as CNN-based ConvNeXt and ViT-based Swin Transformer by a large margin. Pre-trained GC ViT backbones in downstream tasks of object detection, instance segmentation, and semantic segmentation using MS COCO and ADE20K datasets outperform prior work consistently, sometimes by large margins.

EdgeNeXt

Paper	EdgeNeXt: Efficiently Amalgamated CNN-Transformer Architecture for Mobile Vision Applications
Date	21 Jun 2022
Code	https://github.com/mmaaz60/EdgeNeXt
Authors	Muhammad Maaz, Abdelrahman Shaker, Hisham Cholakkal, Salman Khan, Syed Waqas Zamir, Rao Muhammad Anwer, Fahad Shahbaz Khan
Org.	Mohamed Bin Zayed university of Artificial Intelligence, Australian National University, Inception Institute of Artificial Intelligence
Abstract	In the pursuit of achieving ever-increasing accuracy, large and complex neural networks are usually developed. Such models demand high computational resources and therefore cannot be deployed on edge devices. It is of great interest to build resource-efficient general purpose networks due to their usefulness in several application areas. In this work, we strive to effectively combine the strengths of both CNN and Transformer models and propose a new efficient hybrid architecture EdgeNeXt. Specifically in EdgeNeXt, we introduce split depth-wise transpose attention (SDTA) encoder that splits input tensors into multiple channel groups and utilizes depth-wise convolution along with self-attention across channel dimensions to implicitly increase the receptive field and encode multi-scale features. Our extensive experiments on classification, detection and segmentation tasks, reveal the merits of the proposed approach, outperforming state-of-the-art methods with comparatively lower compute requirements. Our EdgeNeXt model with 1.3M parameters achieves 71.2% top-1 accuracy on ImageNet-1K, outperforming MobileViT with an absolute gain of 2.2% with 28% reduction in FLOPs. Further, our EdgeNeXt model with 5.6M parameters achieves 79.4% top-1 accuracy on ImageNet-1K.

2022 Q3

LinGlos(Linear-complexity and Global-local interaction)

Paper	Rethinking Query-Key Pairwise Interactions in Vision Transformers
Date	1 Jul 2022
Code
Authors	Cheng Li, Yangxin Liu
Org.	TCL Research
Abstract	Vision Transformers have achieved state-of-the-art performance in many visual tasks. Due to the quadratic computational and memory complexities of self-attention, recent works either apply attention only to low-resolution inputs or restrict the receptive field to a small local region. To overcome these limitations, we propose key-only attention, which excludes query-key pairwise interactions and uses a compute-efficient saliency-gate to obtain attention weights, modeling local-global interactions in all stages. Key-only attention has linear computational and memory complexities w.r.t input size. We use alternate layout to hybridize convolution and attention layers instead of grafting which is suggested by previous works, so that all stages can benefit from both spatial attentions and convolutions. We leverage these improvements to develop a new self-attention model family, LinGlos, which reach state-of-the-art accuracies on the parameter-limited setting of ImageNet classification benchmark, and outperform baselines significantly in downstream tasks, e.g., COCO object detection and ADE20K semantic segmentation.

UniNet

Paper	UniNet: Unified Architecture Search with Convolution, Transformer, and MLP
Date	12 Jul 2022
Code	https://github.com/Sense-X/UniNet
Authors	Jihao Liu, Xin Huang, Guanglu Song, Yu Liu, Hongsheng Li
Org.	CUHK, MMLab / SenseTime Research
Abstract	Recently, transformer and multi-layer perceptron (MLP) architectures have achieved impressive results on various vision tasks. However, how to effectively combine those operators to form high-performance hybrid visual architectures still remains a challenge. In this work, we study the learnable combination of convolution, transformer, and MLP by proposing a novel unified architecture search approach. Our approach contains two key designs to achieve the search for high-performance networks. First, we model the very different searchable operators in a unified form, and thus enable the operators to be characterized with the same set of configuration parameters. In this way, the overall search space size is significantly reduced, and the total search cost becomes affordable. Second, we propose context-aware downsampling modules (DSMs) to mitigate the gap between the different types of operators. Our proposed DSMs are able to better adapt features from different types of operators, which is important for identifying high-performance hybrid architectures. Finally, we integrate configurable operators and DSMs into a unified search space and search with a Reinforcement Learning-based search algorithm to fully explore the optimal combination of the operators. To this end, we search a baseline network and scale it up to obtain a family of models, named UniNets, which achieve much better accuracy and efficiency than previous ConvNets and Transformers. In particular, our UniNet-B5 achieves 84.9% top-1 accuracy on ImageNet, outperforming EfficientNet-B7 and BoTNet-T7 with 44% and 55% fewer FLOPs respectively. By pretraining on the ImageNet-21K, our UniNet-B6 achieves 87.4%, outperforming Swin-L with 51% fewer FLOPs and 41% fewer parameters.

LightViT

Paper	LightViT: Towards Light-Weight Convolution-Free Vision Transformers
Date	12 Jul 2022
Code	https://github.com/hunto/LightViT
Authors	Tao Huang, Lang Huang, Shan You, Fei Wang, Chen Qian, Chang Xu
Org.	SenseTime Research / The University of Sydney / The University of Tokyo / University of Science and Technology of China
Abstract	Vision transformers (ViTs) are usually considered to be less light-weight than convolutional neural networks (CNNs) due to the lack of inductive bias. Recent works thus resort to convolutions as a plug-and-play module and embed them in various ViT counterparts. In this paper, we argue that the convolutional kernels perform information aggregation to connect all tokens; however, they would be actually unnecessary for light-weight ViTs if this explicit aggregation could function in a more homogeneous way. Inspired by this, we present LightViT as a new family of light-weight ViTs to achieve better accuracy-efficiency balance upon the pure transformer blocks without convolution. Concretely, we introduce a global yet efficient aggregation scheme into both self-attention and feed-forward network (FFN) of ViTs, where additional learnable tokens are introduced to capture global dependencies; and bi-dimensional channel and spatial attentions are imposed over token embeddings. Experiments show that our model achieves significant improvements on image classification, object detection, and semantic segmentation tasks. For example, our LightViT-T achieves 78.7% accuracy on ImageNet with only 0.7G FLOPs, outperforming PVTv2-B0 by 8.2% while 11% faster on GPU.

Dual-ViT

Paper	Dual Vision Transformer
Date	11 Jul 2022
Code	https://github.com/YehLi/ImageNetModel
Authors	Ting Yao, Yehao Li, Yingwei Pan, Yu Wang, Xiao-Ping Zhang, Tao Mei
Org.	JD Explore Academy / Ryerson University
Abstract	Prior works have proposed several strategies to reduce the computational cost of self-attention mechanism. Many of these works consider decomposing the self-attention procedure into regional and local feature extraction procedures that each incurs a much smaller computational complexity. However, regional information is typically only achieved at the expense of undesirable information lost owing to down-sampling. In this paper, we propose a novel Transformer architecture that aims to mitigate the cost issue, named Dual Vision Transformer (Dual-ViT). The new architecture incorporates a critical semantic pathway that can more efficiently compress token vectors into global semantics with reduced order of complexity. Such compressed global semantics then serve as useful prior information in learning finer pixel level details, through another constructed pixel pathway. The semantic pathway and pixel pathway are then integrated together and are jointly trained, spreading the enhanced self-attention information in parallel through both of the pathways. Dual-ViT is henceforth able to reduce the computational complexity without compromising much accuracy. We empirically demonstrate that Dual-ViT provides superior accuracy than SOTA Transformer architectures with reduced training complexity.

Next-ViT

Paper	Next-ViT: Next Generation Vision Transformer for Efficient Deployment in Realistic Industrial Scenarios
Date	12 Jul 2022
Code
Authors	Jiashi Li, Xin Xia, Wei Li, Huixia Li, Xing Wang, Xuefeng Xiao, Rui Wang, Min Zheng, Xin Pan
Org.	ByteDance Inc
Abstract	Due to the complex attention mechanisms and model design, most existing vision Transformers (ViTs) can not perform as efficiently as convolutional neural networks (CNNs) in realistic industrial deployment scenarios, e.g. TensorRT and CoreML. This poses a distinct challenge: Can a visual neural network be designed to infer as fast as CNNs and perform as powerful as ViTs? Recent works have tried to design CNN-Transformer hybrid architectures to address this issue, yet the overall performance of these works is far away from satisfactory. To end these, we propose a next generation vision Transformer for efficient deployment in realistic industrial scenarios, namely Next-ViT, which dominates both CNNs and ViTs from the perspective of latency/accuracy trade-off. In this work, the Next Convolution Block (NCB) and Next Transformer Block (NTB) are respectively developed to capture local and global information with deployment-friendly mechanisms. Then, Next Hybrid Strategy (NHS) is designed to stack NCB and NTB in an efficient hybrid paradigm, which boosts performance in various downstream tasks. Extensive experiments show that Next-ViT significantly outperforms existing CNNs, ViTs and CNN-Transformer hybrid architectures with respect to the latency/accuracy trade-off across various vision tasks. On TensorRT, Next-ViT surpasses ResNet by 5.4 mAP (from 40.4 to 45.8) on COCO detection and 8.2% mIoU (from 38.8% to 47.0%) on ADE20K segmentation under similar latency. Meanwhile, it achieves comparable performance with CSWin, while the inference speed is accelerated by 3.6x. On CoreML, Next-ViT surpasses EfficientFormer by 4.6 mAP (from 42.6 to 47.2) on COCO detection and 3.5% mIoU (from 45.2% to 48.7%) on ADE20K segmentation under similar latency.

XFormer

Paper	Lightweight Vision Transformer with Cross Feature Attention
Date	15 Jul 2022
Code
Authors	Youpeng Zhao, Huadong Tang, Yingying Jiang, Yong A, Qiang Wu
Org.	Samsung Research China / University of Technology Sydney
Abstract	Recent advances in vision transformers (ViTs) have achieved great performance in visual recognition tasks. Convolutional neural networks (CNNs) exploit spatial inductive bias to learn visual representations, but these networks are spatially local. ViTs can learn global representations with their self-attention mechanism, but they are usually heavy-weight and unsuitable for mobile devices. In this paper, we propose cross feature attention (XFA) to bring down computation cost for transformers, and combine efficient mobile CNNs to form a novel efficient light-weight CNN-ViT hybrid model, XFormer, which can serve as a general-purpose backbone to learn both global and local representation. Experimental results show that XFormer outperforms numerous CNN and ViT-based models across different tasks and datasets. On ImageNet1K dataset, XFormer achieves top-1 accuracy of 78.5% with 5.5 million parameters, which is 2.2% and 6.3% more accurate than EfficientNet-B0 (CNN-based) and DeiT (ViT-based) for similar number of parameters. Our model also performs well when transferring to object detection and semantic segmentation tasks. On MS COCO dataset, XFormer exceeds MobileNetV2 by 10.5 AP (22.7 -> 33.2 AP) in YOLOv3 framework with only 6.3M parameters and 3.8G FLOPs. On Cityscapes dataset, with only a simple all-MLP decoder, XFormer achieves mIoU of 78.5 and FPS of 15.3, surpassing state-of-the-art lightweight segmentation networks.

CETNet

Paper	Convolutional Embedding Makes Hierarchical Vision Transformer Stronger
Date	27 Jul 2022
Code
Authors	Cong Wang, Hongmin Xu, Xiong Zhang, Li Wang, Zhitong Zheng, Haifeng Liu
Org.	OPPO / North China University of Technology / University of Science and Technology of China / Neolix Autonomous Vehicle
Abstract	Vision Transformers (ViTs) have recently dominated a range of computer vision tasks, yet it suffers from low training data efficiency and inferior local semantic representation capability without appropriate inductive bias. Convolutional neural networks (CNNs) inherently capture regional-aware semantics, inspiring researchers to introduce CNNs back into the architecture of the ViTs to provide desirable inductive bias for ViTs. However, is the locality achieved by the micro-level CNNs embedded in ViTs good enough? In this paper, we investigate the problem by profoundly exploring how the macro architecture of the hybrid CNNs/ViTs enhances the performances of hierarchical ViTs. Particularly, we study the role of token embedding layers, alias convolutional embedding (CE), and systemically reveal how CE injects desirable inductive bias in ViTs. Besides, we apply the optimal CE configuration to 4 recently released state-of-the-art ViTs, effectively boosting the corresponding performances. Finally, a family of efficient hybrid CNNs/ViTs, dubbed CETNets, are released, which may serve as generic vision backbones. Specifically, CETNets achieve 84.9% Top-1 accuracy on ImageNet-1K (training from scratch), 48.6% box mAP on the COCO benchmark, and 51.6% mIoU on the ADE20K, substantially improving the performances of the corresponding state-of-the-art baselines.

XFormer

Paper	Lightweight Vision Transformer with Cross Feature Attention
Date	15 Jul 2022
Code
Authors	Youpeng Zhao, Huadong Tang, Yingying Jiang, Yong A, Qiang Wu
Org.	Samsung Research China / University of Technology Sydney
Abstract	Recent advances in vision transformers (ViTs) have achieved great performance in visual recognition tasks. Convolutional neural networks (CNNs) exploit spatial inductive bias to learn visual representations, but these networks are spatially local. ViTs can learn global representations with their self-attention mechanism, but they are usually heavy-weight and unsuitable for mobile devices. In this paper, we propose cross feature attention (XFA) to bring down computation cost for transformers, and combine efficient mobile CNNs to form a novel efficient light-weight CNN-ViT hybrid model, XFormer, which can serve as a general-purpose backbone to learn both global and local representation. Experimental results show that XFormer outperforms numerous CNN and ViT-based models across different tasks and datasets. On ImageNet1K dataset, XFormer achieves top-1 accuracy of 78.5% with 5.5 million parameters, which is 2.2% and 6.3% more accurate than EfficientNet-B0 (CNN-based) and DeiT (ViT-based) for similar number of parameters. Our model also performs well when transferring to object detection and semantic segmentation tasks. On MS COCO dataset, XFormer exceeds MobileNetV2 by 10.5 AP (22.7 -> 33.2 AP) in YOLOv3 framework with only 6.3M parameters and 3.8G FLOPs. On Cityscapes dataset, with only a simple all-MLP decoder, XFormer achieves mIoU of 78.5 and FPS of 15.3, surpassing state-of-the-art lightweight segmentation networks.

HorNet

Paper	HorNet: Efficient High-Order Spatial Interactions with Recursive Gated Convolutions
Date	28 Jul 2022
Code	https://github.com/raoyongming/HorNet
Authors	Yongming Rao, Wenliang Zhao, Yansong Tang, Jie Zhou, Ser-Nam Lim, Jiwen Lu
Org.	Tsinghua University / Meta AI
Abstract	Recent progress in vision Transformers exhibits great success in various tasks driven by the new spatial modeling mechanism based on dot-product self-attention. In this paper, we show that the key ingredients behind the vision Transformers, namely input-adaptive, long-range and high-order spatial interactions, can also be efficiently implemented with a convolution-based framework. We present the Recursive Gated Convolution (gnConv) that performs high-order spatial interactions with gated convolutions and recursive designs. The new operation is highly flexible and customizable, which is compatible with various variants of convolution and extends the two-order interactions in self-attention to arbitrary orders without introducing significant extra computation. gnConv can serve as a plug-and-play module to improve various vision Transformers and convolution-based models. Based on the operation, we construct a new family of generic vision backbones named HorNet. Extensive experiments on ImageNet classification, COCO object detection and ADE20K semantic segmentation show HorNet outperform Swin Transformers and ConvNeXt by a significant margin with similar overall architecture and training configurations. HorNet also shows favorable scalability to more training data and a larger model size. Apart from the effectiveness in visual encoders, we also show gnConv can be applied to task-specific decoders and consistently improve dense prediction performance with less computation. Our results demonstrate that gnConv can be a new basic module for visual modeling that effectively combines the merits of both vision Transformers and CNNs.

Token Fusion

Paper	Improved Image Classification with Token Fusion
Date	19 Aug 2022
Code
Authors	Keong Hun Choi, Jin Woo Kim, Yao Wang, Jong Eun Ha
Org.	Seoul National University of Science and Technology / Ways1 Inc.
Abstract	In this paper, we propose a method using the fusion of CNN and transformer structure to improve image classification performance. In the case of CNN, information about a local area on an image can be extracted well, but there is a limit to the extraction of global information. On the other hand, the transformer has an advantage in relatively global extraction, but has a disadvantage in that it requires a lot of memory for local feature value extraction. In the case of an image, it is converted into a feature map through CNN, and each feature map's pixel is considered a token. At the same time, the image is divided into patch areas and then fused with the transformer method that views them as tokens. For the fusion of tokens with two different characteristics, we propose three methods: (1) late token fusion with parallel structure, (2) early token fusion, (3) token fusion in a layer by layer. In an experiment using ImageNet 1k, the proposed method shows the best classification performance.