Segmentation evaluation database

The goal of this work is to provide an empirical and scientific basis for research on image segmentation. Evaluating the results produced by segmentation algorithms is challenging, as it is difficult to come up with canonical test sets providing ground truth segmentations. This is partly because manual delineation of segments in everyday complex images can be laborious. Furthermore, people often tend to incorporate into their segmentations semantic considerations which are beyond the scope of data driven segmentation algorithms. For this reason many existing algorithms show only few segmentation results. To evaluate the segmentation produced by different algorithms we have compiled a database, currently containing 200 gray level images along with ground truth segmentations. The database is specially designed to avoid potential ambiguities by only incorporating images that clearly depict one or two object/s in the foreground that differ from its surroundings by either intensity, texture, or other low level cues. The ground truth segmentation were obtained by asking human subjects to manually segment the gray scale images (the color source is also provided) into two or three classes with each image segmented by three different human subjects. The segmentation is evaluated by assessing its consistency with the ground truth segmentation and their amounts of fragmentation. Together with this database evaluation we have provided a code for the evaluation of the given segmentation algorithm. That way different segmentation algorithm may have comparable results for more details see the “Evaluation tests” section. If you use this database you agree to the disclaimer below and include a citation to our CVPR 2007 paper:

@inproceedings{AlpertGBB07,
   author = {"Sharon Alpert and Meirav Galun and Ronen Basri and Achi Brandt"},
   title = {"Image Segmentation by Probabilistic Bottom-Up Aggregation and Cue                             Integration."},
   booktitle = {"Proceedings of the IEEE Conference on Computer Vision and Pattern                                 Recognition"},
   month = {"June"},
   year = {"2007"}
}

Disclaimer

This database is made available for research purposes only. The images were obtained a few royalty free images databases, the source is indicated near each image, without permission from the original copyright holders. By downloading these files, you agree not to hold the authors or The Weizmann institute of science liable for any damage, lawsuits, or other loss resulting from the possession or use of these images and/or ground truth files. You also acknowledge that you will act according to the terms of use of each image as specified on its source site. We reserve the right to change the database at any time without notice. If you are the copyright owner of one of these images and would like it removed from the dataset, please use the contact at the bottom of the page.






This page is maintained by Sharon Alpert