Contact: Daniel Kane
dbkane@ucsd.edu
858-534-3262
University of California - San Diego

Context Fixes Image Labeling Problem

Caption: Thanks to context information, an inaccurate cow label is correctly turned into a boat. Left to Right: Original segmented image; Categorization without contextual contraints; Categorization with co-occurence contextual constraints derived from the training data; Ground truth.

Credit: UC San Diego

Usage Restrictions: Mandatory credit: UC San Diego

Related news release: UC San Diego researchers give computers 'common sense'