2
2

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.

【Mono Depth】L1 LossよりReprojection Lossが良いらしい

Posted at

Robust Semi-Supervised Monocular Depth Estimation with Reprojected Distances

![image.png](https://qiita-image-store.s3.ap-northeast-1.amazonaws.com/0/482094/8a8f9cb8-848e-2501-fb8c-969649981b83.png) この論文ではReprojection LossというLossを新たに提案している。

アルゴリズム

![image.png](https://qiita-image-store.s3.ap-northeast-1.amazonaws.com/0/482094/b92f0589-4c3b-7220-be9f-05ea18aefc5f.png) Lphoto => per-pixel photometric loss Mphoto => カメラと同じ速度で動いている物体とカメラが動いていない時のマスク Lsmooth =>Depthの2回微分 Lrep =>今回提案されるreprojection loss

新規性

![image.png](https://qiita-image-store.s3.ap-northeast-1.amazonaws.com/0/482094/c163b179-26b5-da75-47d3-e46c9cbbde51.png) LiDARを使ったSupervise LossはDepth Mapの距離を比較するのだが、 Lrepでは"推定したDepth"と"Ground TruthのDepth"をPoseNetによって推定されたTt->sによってwarpさせる warpした時にDepthが違えばReprojectionされる位置がずれる。その差分をlossとして定義する。

image.png
数式で表すとこうなるらしい。

結果

![image.png](https://qiita-image-store.s3.ap-northeast-1.amazonaws.com/0/482094/d19d7661-32dd-22e7-f682-b3ee01fc4908.png) DepthのLossをL1よりReprojection lossにしたほうが精度が出たらしい。

結論

・なぜだか理解できていないが、L1 LossよりReprojection Lossの方が精度が上がるらしい。

参考文献

Robust Semi-Supervised Monocular Depth Estimation with Reprojected Distances https://arxiv.org/pdf/1910.01765.pdf
2
2
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
2
2

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?