This is an evolving piece of code that implements the techniques in this paper:

@InProceedings{BharathICCV2011,
  author       = "Bharath Hariharan and Pablo Arbelaez and Lubomir Bourdev and Subhransu Maji and Jitendra Malik",
  title        = "Semantic Contours from Inverse Detectors",
  booktitle    = "International Conference on Computer Vision (ICCV)",
  year         = "2011",
}


Currently, the repo contains code for bootstrapping an ``inverse detector" out of a HOG detector(in the folder ``inverse detector") and code for using the ``inverse detector" to reweight contours (in the folder ``reweight"). The file demo.m is a readable demothat explains how to run the code.

The code for reweighting can also be used for other things, such as for reweighting pixel values based on segmentation masks associated with detections etc.