PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume

License

Copyright (C) 2018 NVIDIA Corporation. All rights reserved. Licensed under the CC BY-NC-SA 4.0 license (https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode).

Usage

For Caffe users, please refer to Caffe/README.md.

For PyTorch users, please refer to PyTorch/README.md

Note that, currently, the PyTorch implementation is inferior to the Caffe implementation (~3% performance drop on Sintel). These are due to differences in implementation between Caffe and PyTorch, such as image resizing and I/O.

Network Architecture

PWC-Net fuses several classic optical flow estimation techniques, including image pyramid, warping, and cost volume, in an end-to-end trainable deep neural networks for achieving state-of-the-art results.

Paper & Citation

Deqing Sun, Xiaodong Yang, Ming-Yu Liu, and Jan Kautz. "PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume." CVPR 2018 or arXiv:1709.02371

Project page link

Talk at robust vision challenge workshop

Talk at CVPR 2018 conference (starts ~7:00)

If you use PWC-Net, please cite the following paper:

@InProceedings{Sun2018PWC-Net,
  author    = {Deqing Sun and Xiaodong Yang and Ming-Yu Liu and Jan Kautz},
  title     = {{PWC-Net}: {CNNs} for Optical Flow Using Pyramid, Warping, and Cost Volume},
  booktitle = CVPR,
  year      = {2018},
}

or the arXiv paper

@article{sun2017pwc,
  author={Sun, Deqing and Yang, Xiaodong and Liu, Ming-Yu and Kautz, Jan},
  title={{PWC-Net}: {CNNs} for Optical Flow Using Pyramid, Warping, and Cost Volume},
  journal={arXiv preprint arXiv:1709.02371},
  year={2017}
}

Related Work from NVIDIA

flownet2-pytorch

Contact

Deqing Sun ([email protected])