Tools to extract dense optical flow from videos, based on OpenCV
Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is 40 commits ahead, 2 commits behind wanglimin:master.
Latest commit c9369a3 Jul 29, 2017

Extracting dense flow field given a video.


  • LibZip: to install on ubuntu apt-get install libzip-dev on mac brew install libzip

For OpenCV 3 Users

Please see the opencv-3.1 branch. Many thanks to @victorhcm for the contributions!


git clone --recursive
mkdir build && cd build
cmake .. && make -j


./extract_gpu -f test.avi -x tmp/flow_x -y tmp/flow_y -i tmp/image -b 20 -t 1 -d 0 -s 1 -o dir
  • test.avi: input video
  • tmp: folder containing RGB images and optical flow images
  • dir: output generated images to folder. if set to zip, will write images to zip files instead.

Warp Flow

The warp optical flow is used in the following paper

  author    = {Limin Wang and
               Yuanjun Xiong and
               Zhe Wang and
               Yu Qiao and
               Dahua Lin and
               Xiaoou Tang and
               Luc {Van Gool}},
  title     = {Temporal Segment Networks: Towards Good Practices for Deep Action Recognition},
  booktitle   = {ECCV},
  year      = {2016},

To extract warp flow, use the command

./extract_warp_gpu -f test.avi -x tmp/flow_x -y tmp/flow_y -i tmp/image -b 20 -t 1 -d 0 -s 1 -o dir