search

NeurAI-Lab / biHomE

This is the official repo for the CVPR 2021 IMW paper: "Perceptual Loss for Robust Unsupervised Homography Estimation"
MIT License
37 stars 5 forks source link

Some question about RANSAC #9

Closed Acmenwangtuo closed 2 years ago

Acmenwangtuo commented 2 years ago

Hello,in PFNet ,we should use ransac to sample point from PF, how u implement it , i use the kornia the cuda out of memory

dkoguciuk commented 2 years ago

Hi @Acmenwangtuo ,

exactly as you say, For training, we don't really need the real value of homography, so I only return delta_hat as a prediction of corner points of the image: https://github.com/NeurAI-Lab/biHomE/blob/main/src/heads/NoOpHead.py#L34

However, we need to use some method to find the homography out of the perspective field for testing, and here comes the RANSAC algorithm: https://github.com/NeurAI-Lab/biHomE/blob/main/src/heads/NoOpHead.py#L76

In general, the Zeng method takes much more GPU memory than other models implemented in this repo.

Best, Daniel

dkoguciuk commented 2 years ago

Hi @Acmenwangtuo ,

I am closing the issue for now. Feel free to reopen if you'd like.

Best, Daniel