Summary
This guide outlines the setup process for ViTPose, a state-of-the-art 2D/3D pose estimation model, on a Windows 11 system with an RTX 4080 GPU. While the official documentation primarily supports Linux, this setup demonstrates a successful installation on Windows.
Device Configuration
- Operating System: Windows 11 (Official support: Linux)
- GPU: NVIDIA RTX 4080
- NVIDIA Driver: 566.07
- CUDA Version: 11.1 (Download CUDA 11.1.1)
- cuDNN Version: 8.0.5 (Download cuDNN 8.0.5)
- Development Tools:
- MSYS2 + MinGW-w64
- Python: 3.7.16
- Visual Studio: 2019
- GCC: 14.2.0 (Rev2, Built by MSYS2 project).
- System PATH: MinGW-w64 is added to Windows' system PATH
| Library | Version |
|---|---|
| PyTorch | 1.9.0 + cu111 |
| TorchVision | 0.10.0 + cu111 |
| MMCV (mmcv-full) | 1.3.9 |
| MMPose | 0.24.0 |
| OpenCV | 4.11.0 |
| xtcocotools | 1.14.3 |
| timm | 0.4.9 |
| scipy | 1.7.3 |
| openmim | 0.3.9 |
| opencv-python | 4.11.0.86 |
| opencv-contrib-python | 4.11.0.86 |
| numpy | 1.21.6 |
| matplotlib | 3.5.3 |
| json-tricks | 3.17.3 |
Installation Steps
conda create -n vitpose_official python=3.7.6 -y
conda activate vitpose_test
conda install pytorch==1.9.0 torchvision torchaudio cudatoolkit=11.1 -c pytorch -c nvidia
pip install timm==0.4.9 einops
pip install matplotlib==3.5.3
pip install scipy==1.7.3
pip install json-tricks==3.17.3
pip install numpy==1.21.6
% clone ViTPose
git clone https://github.com/ViTAE-Transformer/ViTPose.git
cd ViTPose
git clone https://github.com/open-mmlab/mmcv.git
cd mmcv
git checkout v1.3.9
set MMCV_WITH_OPS=1
pip install -e .
======
(maybe below code also okay)
pip install -U openmim
mim install mmcv-full==1.3.9
=====
cd ..
git clone https://github.com/open-mmlab/mmpose.git
cd mmpose
pip install -r requirements.txt
pip install -v -e .
cd ..
pip install -v -e .