1
1

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?

More than 3 years have passed since last update.

【semantic segmentation】Fully Convolutional Networksを理解してみる

Last updated at Posted at 2020-11-12

概要

Fully Convolutional Networkを理解してsemantic segmentationを理解してみる

Base Network

![image.png](https://qiita-image-store.s3.ap-northeast-1.amazonaws.com/0/482094/ba1f76c7-855b-4014-56cf-c3304cf50218.png)

図からみて分かるようにFCNはConvolutionのみを行い最終的にupsample(interpolation)する。
よって出力のsemantic mapの解像度は低くなる

改善

image.png

前の情報を足し合わせる事で精度を上げている

FC-32s
1列目はconv7の結果を32倍upsampleしてsegmentation mapを得る

FC-16s
2列目はconv7の結果を2倍upsampleしたものとpool4の結果を足し合わせて、16倍upsampleしてsegmentation mapを得る

FC-8s
3列目はconv7の結果を4倍upsampleしたものとpool4の結果を2倍upsampleしたものとpool3の結果を足し合わせてたものに、8倍upsampleしてsegmentation mapを得る

結果

![image.png](https://qiita-image-store.s3.ap-northeast-1.amazonaws.com/0/482094/cade2d72-cd3a-5397-5ef5-5f63932b33de.png)

図から見てわかるようにFC-32s->FC-16s->FC-8sになるに連れて精度が上がっている事が分かる。

FCNの特徴はDecordしない所かな!

参考文献

Fully Convolutional Networks for Semantic Segmentation https://arxiv.org/pdf/1605.06211.pdf https://github.com/wkentaro/pytorch-fcn/blob/master/torchfcn/models/fcn8s.py
1
1
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
1
1

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?