From Region Features to Semantic Labels: a Probabilistic Approach.
Rui Li, Wee Kheng Leow · 2003
This paper presents a novel method of semantic labeling that can assign multiple semantic labels to a region along with the con dence measures of the assignment. Unlike existing classi cation methods, it can learn to perform semantic labeling incrementally. Test results show that the method is eective and accurate in labeling a wide variety of regions