Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Rethinking Performance Gains in Image Dehazing Networks #35

Open
IsHYuhi opened this issue Feb 4, 2023 · 0 comments
Open

Rethinking Performance Gains in Image Dehazing Networks #35

IsHYuhi opened this issue Feb 4, 2023 · 0 comments
Assignees
Labels

Comments

@IsHYuhi
Copy link
Owner

IsHYuhi commented Feb 4, 2023

INFO

author

Yuda Song† Yang Zhou† Hui Qian Xin Du

affiliation

Zhejiang University, Hangzhou, China

conference or year

2022

link

arXiv
実装

概要

スクリーンショット 2023-02-04 12 03 28

dehazingタスクに既存の研究で提案されたネットワークやパイプラインは問題なく動作するが,キーとなるメカニズムが依然として不明瞭である.この研究では複雑なモジュール等を用いることなく最小限の修正をUNetに加えることで,コンパクトなdehazingネットワークを提案.複数のdehazingデータセットにおいて他のSoTA手法に勝ることをを示した.

提案手法

スクリーンショット 2023-02-04 12 05 21

gConv: ゲート機構の畳み込み.Depth-wiseとPoint-wiseを用いている.NAFNetに近い構造だが,オリジナルのGLUに則って非線形活性化関数にシグモイドを用いている.
SK Fusion:局所と大域の特徴マップをチャンネルアテンション(SE-Layer)を通した後に加算してるだけ.

検証

スクリーンショット 2023-02-04 12 05 30

スクリーンショット 2023-02-04 12 05 34

ablation study

スクリーンショット 2023-02-04 12 06 07

新規性

議論,展望

Comment

Ablation Studyからわかるように最適な学習方法やレイヤーの選択を行なっている.それによって軽くてシンプルなネットワークを実現している.
単純に精度の観点から言えば,先行研究も同様に最適なハイパラの選択を行えば,この研究を超えるような精度もでる気がする.

PixelShuffleとUpsamplingのablationも見たかった.
LayerNormは空間相関を破壊するとDehazeFormerで述べられているが,InstanceNormがBatchNormより大幅に下がる理由はなんだろう.

date

Feb. 4th 2023

@IsHYuhi IsHYuhi self-assigned this Feb 4, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

No branches or pull requests

1 participant