Fork me on GitHub

Deep Extreme Cut (DEXTR):
From Extreme Points to Object Segmentation

State of the Art in guided and interactive segmentation


K.K. Maninis*, S. Caelles*, J. Pont-Tuset, and L. Van Gool
Deep Extreme Cut: From Extreme Points to Object Segmentation,
Computer Vision and Pattern Recognition (CVPR), 2018.
[BibTeX] [PDF]
  author 	= {K.-K. Maninis and S. Caelles and J. Pont-Tuset and L. {Van Gool}},
  title 	= {Deep Extreme Cut: From Extreme Points to Object Segmentation},
  booktitle	= {Computer Vision and Pattern Recognition (CVPR)},
  year 		= {2018}


The table shows the overall results of DEXTR, compared to the state-of-the-art interactive segmentation methods. Using only 4 extreme clicks, we obtain top-quality segmentations.

Below the quality per annotation budget, using DEXTR for annotating PASCAL, and PSPNet to train for semantic segmentation.
Qualitative Results of DEXTR