Awesome

Correspondence-wise Losses

[Arxiv] | [Website]

Comparing Correspondences: Video Prediction with Correspondence-wise Losses Daniel Geng, Max Hamilton, Andrew Owens CVPR, 2022

Teaser figure

We propose correspondence-wise losses, which align images before calculating an arbitrary loss. In our paper we show that this results in improvements to image synthesis tasks and provides robustness to spatial uncertainty.

Code

Code and additional information can be found in the code directory.

Citation

If you found this code useful please consider citing our paper:

@inproceedings{geng2022correspondences,
  title={Comparing Correspondences: Video Prediction with Correspondence-wise Losses},
  author={Geng, Daniel and Hamilton, Max and Owens, Andrew},
  booktitle={CVPR},
  year={2022}
}

Acknowledgements

DG is supported by a National Science Foundation Graduate Research Fellowship under Grant No. 1841052.