This repository contains implementation code for the project 'End-to-End Learning of Motion Representation for Video Understanding' (CVPR 2018).

http://lijiefan.me/project_webpage/TVNet_cvpr/index.html

We use tensorflow (https://www.tensorflow.org) for our implementation.

We use .mat file for TVNet generated results saving, and Matlab for results visualization.

Our current release has been tested on Ubuntu 16.04.

git clone https://github.com/LijieFan/tvnet.git

The file (demo.py) has the following options:

• -scale: Number of scales in TVNet (default: 1)
• -warp: Number of warppings in TVNet (default: 1)
• -iteration: Number of iterations in TVNet(default: 50)
• -gpu: the gpu to run on (0-indexed, -1 for CPU)

Sample usages include

• Generate motion representation for frames in frame/img1.png and frame/img2.png.

python demo.py --scale 1 --warp 1 --iteration 50 --gpu 1

-TVNet generated results are saved in result/result.mat

-Use the MPI-Sintel tool box for result visualization. In matlab, run run visualize/visualize.m.

We’d love to express out appreciation to Jian Guo for the useful discussions during the course of this research.

if you find our code useful for your research, please cite our paper:

@inproceedings{fan2018end,
title={End-to-End Learning of Motion Representation for Video Understanding},
author={Fan, Lijie and Huang, Wenbing and Gan, Chuang and Ermon, Stefano and Gong, Boqing and Huang, Junzhou},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
pages={},
year={2018}
}