0
0

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.

【Stereo Depth】SCV-Net : Sparse Cost Volume Network

Posted at

Sparse Cost Volume for Efficient Stereo Matching

![image.png](https://qiita-image-store.s3.ap-northeast-1.amazonaws.com/0/482094/1f80874d-402b-24f4-2fce-8fba5572c53e.png) Cost Volumeを作る代表的なモデルとして[GC-Net](https://qiita.com/minh33/items/8f3ce0ad64035d994af6)があるが、計算速度が遅い。 この論文では計算速度向上の為Cost Volumeの作り方を少し工夫している。

新規性

Sparse Cost Volume Construction

![image.png](https://qiita-image-store.s3.ap-northeast-1.amazonaws.com/0/482094/1ae1d237-3fb2-2f61-25e6-084cd867fbc3.png)

cost volumeを作る時通常は1pixelずつスライドさせて(Batch,Disparity,Channel,Height,Width)作るのだが、SCV-Netでは
3pixelずつスライドさせて(Batch,Disparity/3,Channel,Height,Width)のコストボリュームを作る

もちろん計算速度が大幅に上がる。

結論

・Cost Volumeを減らし、layerを増やすなどの他のoperationを足す事で、計算コストを削減しつつ、同等の精度が出る。

ずっと思ってたけど、Cost Volumeってかなり無駄が多い気がする。

参考文献

Sparse Cost Volume for Efficient Stereo Matching https://www.researchgate.net/publication/329134120_Sparse_Cost_Volume_for_Efficient_Stereo_Matching
0
0
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
0
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?