Join GitHub today
GitHub is home to over 50 million developers working together to host and review code, manage projects, and build software together.Sign up
GitHub is where the world builds software
Millions of developers and companies build, ship, and maintain their software on GitHub — the largest and most advanced development platform in the world.
Perceptually Inspired Layout-aware Losses for Image Segmentation ================================================================ This software implements the method described in the following paper: A. Osokin and P. Kohli Perceptually Inspired Layout-aware Losses for Image Segmentation In European Conference on Computer Vision (ECCV), 2014. The paper can be downloaded from: http://bayesgroup.ru/wp-content/uploads/2014/07/skeletalLossesLearning_eccv2014_cameraReady.pdf If you use the software in your work please cite the aforementioned paper. 1. Installation -------------------- 1) Run setup.m to add the required folders to Matlab Path 2) Run compile_mex.m to compile all the MEX-functions on your system. Tested with Matlab R2014a + MSVC 2012 on Win7 and Matlab R2014b + gcc 4.6 on Ubuntu 12.04 3) Download the dataset provided by the following paper: V. Gulshan, C. Rother, A. Criminisi, A. Blake and A. Zisserman Geodesic Star Convexity for Interactive Image Segmentation In Computer Vision and Pattern Recognition (CVPR), 2010. Images: http://www.robots.ox.ac.uk/~vgg/data/iseg/data/images.tgz Ground truth: http://www.robots.ox.ac.uk/~vgg/data/iseg/data/images-gt.tgz Brush strokes: http://www.robots.ox.ac.uk/~vgg/data/iseg/data/images-labels.tgz Unpack everything to ./data 4) Run ./experiments_eccv2014/example_training.m to test how everything is working Run ./experiments_eccv2014/example_motivation.m to see Table 1 of the paper 5) Run ./experiments_eccv2014/run_full_experiment.m to reproduce the ECCV 2014 experiments Note, the full experiment requires 280 runs of the training procedure which will take quite some time 2. Third-party software -------------------- This software uses several third-party packages (some of the packages are released under the research-only license). Please, consider citing the corresponding papers: 1) IBFS algorithm to solve the max-flow/min-cut problem: http://www.cs.tau.ac.il/~sagihed/ibfs/ A. V. Goldberg, S. Hed, H. Kaplan, R. E. Tarjan, and R. F. Werneck, Maximum Flows by Incremental Breadth-First Search, In Proceedings of the 19th European conference on Algorithms, ESA'11, pages 457-468. 2) Geodesic star convexity http://www.robots.ox.ac.uk/~vgg/software/iseg/ V. Gulshan, C. Rother, A. Criminisi, A. Blake and A. Zisserman, Geodesic star convexity for interactive image segmentation. In Proceedings of Conference on Vision and Pattern Recognition (CVPR 2010). 3) Learning Low-order Models for Enforcing High-order Statistics https://github.com/ppletscher/hol P. Pletscher and P. Kohli Learning Low-order Models for Enforcing High-order Statistics AISTATS, 2012.