difficult segmentation data code

Here (.zip .tar.gz) is a version of the C++ code for segmenting images with cooperative cuts, including an example. The code assumes that you have the types/groups of edges ready. In the paper, we used k-means clustering, where as features we used either the RGB gradient along an edge, or the intensity ratio of its adjacent pixels. The code for generating those is available upon request to the authors. See the README for details.

If you use this code or build on it, please cite the associated paper:

S. Jegelka and J. Bilmes. Submodularity beyond submodular energies: coupling edges in graph cuts. CVPR 2011. (here is the supplementary material)

The data for the experiments in the paper is available here.