Supervised Evaluation of Image Segmentation

Resource Type Date
Software 2013-06-24

Description

 

UPDATE: The new version of the project page and code can be found here: http://vision.ee.ethz.ch/~cvlsegmentation/seism/. Check it out!

##########################################################################################

Code package that implements the image segmentation measures and reproduces all results from the CVPR2013 paper:

Measures and Meta-Measures for the Supervised Evaluation of Image Segmentation
Jordi Pont-Tuset and Ferran Marques, CVPR 2013.

This package allows you to easily evaluate your segmentation method using the precision-recall for boundaries and the new precision-recall for objects and parts:

 

To download the package click on the icon (or github page) and follow the instructions in the readme file:

To fully reproduce the results from scratch, you need to download the partitions obtained by six state-of-the-art segmentation methods and put them in a folder "datasets" inside the SEISM folder. Click on the icon to download these datasets (572MB):

Specifically, the dataset contains the following pre-computed partitions on the full BSDS500 segmentation database:

  • Ultrametric Contour Maps (Arbelaez et al.TPAMI 2011)
  • Normalized Cuts (Shi & Malik, TPAMI 2000)
  • Mean Shift (Comaniciu & Meer, TPAMI 2002)
  • Efficient Graph-Based Image Segmentation (Felzenszwalb & Huttenlocher, IJCV 2004)
  • NWMC Binary Partition Tree (Vilaplana et al. TIP 2008)
  • IID-KL Binary Partition Tree (Calderero & Marques, TIP 2010)

 

Enjoy! And please cite the paper if you use the code or datasets.

 

 

People involved

Jordi Pont-Tuset PhD Candidate
Ferran Marqués Professor

Related Publications

J. Pont-Tuset, Image Segmentation Evaluation and Its Application to Object Detection, Universitat Politècnica de Catalunya, BarcelonaTech, Barcelona, 2014. (48.44 MB)
J. Pont-Tuset and Marqués, F., Measures and Meta-Measures for the Supervised Evaluation of Image Segmentation, in Computer Vision and Pattern Recognition (CVPR), 2013. (909.27 KB)