Learning to Detect Motion Boundaries

Philippe Weinzaepfel Jerome Revaud Zaid Harchaoui Cordelia Schmid

YouTube Motion Boundaries (YMB) dataset

The YMB dataset can be donwloaded here.

Motion boundaries detection code

Please note that our code is mentioned only for scientific or personal use and is published under the MSR-LA Full Right license.

Download

The detection code can be downloaded here.

Installation

This code is written for the Matlab interpreter (tested with versions R2014a-2014b) and requires the Matlab Image Processing Toolbox.
Additionally, Piotr's Matlab Toolbox (version 3.26 or later) written by Piotr Dollar is also required.
Next, please compile mex code from within Matlab:
- mex private/edgesDetectMex.cpp -outdir private [OMPPARAMS]
- mex private/image_warping_error.cpp -outdir private [OMPPARAMS]
Here [OMPPARAMS] are parameters for OpenMP and are OS and compiler dependent.
- Windows: [OMPPARAMS] = '-DUSEOMP' 'OPTIMFLAGS="$OPTIMFLAGS' '/openmp"'
- Linux: [OMPPARAMS] = '-DUSEOMP' CXXFLAGS="\$CXXFLAGS -fopenmp" LDFLAGS="\$LDFLAGS -fopenmp"
To compile without OpenMP simply omit [OMPPARAMS]; note that code will be single threaded in this case.
Models based on Classic+NL-fast optical flow estimation are provided.
Models based on LDOF optical flow estimation can be downloaded here.

Getting started

Make sure to carefully follow the installation instructions above and to have Piotr's Matlab Toolbox in your path.
Run demo.m for example usage.

History

Version 1.0 (June 2015)

Citation

If you use our code or the YMB dataset, please cite our CVPR'15 paper:

@inproceedings{weinzaepfel:hal-01142653,
  TITLE = {{Learning to Detect Motion Boundaries}},
  AUTHOR = {Weinzaepfel, Philippe and Revaud, Jerome and Harchaoui, Zaid and Schmid, Cordelia},
  URL = {https://hal.inria.fr/hal-01142653},
  BOOKTITLE = {{CVPR 2015 - IEEE Conference on Computer Vision \& Pattern Recognition}},
  ADDRESS = {Boston, United States},
  YEAR = {2015},
  MONTH = Jun,
}