README.md

FlowNet2 (PyTorch v0.3.0)

Pytorch implementation of FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks. Most part are from this repo, we made it as a off-the-shelf package:

• After installation, just copy the whole folder FlowNet2_src to your codebase to use. See demo.py for details.

Environment

This code has been test with Python3.6 and PyTorch0.3.0, with a Tesla K80 GPU. The system is Ubuntu 14.04.

Installation

# install custom layers
cd FlowNet2_src
bash install.sh

Note: you might need to modify here, here, and here, according to the GPU you use.

Converted Caffe Pre-trained Models

• FlowNet2[620MB]
• FlowNet2-C[149MB]
• FlowNet2-CS[297MB]
• FlowNet2-CSS[445MB]
• FlowNet2-CSS-ft-sd[445MB]
• FlowNet2-S[148MB]
• FlowNet2-SD[173MB]

Inference mode

First download pre-trained models of FlowNet2 and modify the path, then

python demo.py

If installation is sucessful, you should see the following:

Reference

If you find this implementation useful in your work, please acknowledge it appropriately and cite the paper using:

@InProceedings{IMKDB17,
  author       = "E. Ilg and N. Mayer and T. Saikia and M. Keuper and A. Dosovitskiy and T. Brox",
  title        = "FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks",
  booktitle    = "IEEE Conference on Computer Vision and Pattern Recognition (CVPR)",
  month        = "Jul",
  year         = "2017",
  url          = "http://lmb.informatik.uni-freiburg.de//Publications/2017/IMKDB17"
}

Acknowledgments

• NVIDIA/flownet2-pytorch: Most part
• hellock/flownet2-pytorch: Python3.x and PyTorch0.3.0 support