This preview shows page 1. Sign up to view the full content.
Unformatted text preview: Evaluating multi-class learning strategies in a hierarchical framework for object detection. Sanja Fidler, Marko Boben, Aleš Leonardis. http://vicos.fri.uni-lj.si/ Summary Objects are represented with a hierarchical compositional shape vocabulary which is learned from images. We evaluate 3 different strategies for learning a hierarchical multi-class object vocabulary for object detection: independent, joint and sequential training. We explore and compare their computational behavior (space and time) and detection performance as a function of the number of learned object classes on several recognition datasets. Results
10 classes. 50 classes.
Detections Sharing between visually similar and dissimilar classes. 2 classes. Ind. Joint Seq. Joint ...
View Full Document
This note was uploaded on 02/12/2010 for the course COMPUTER S 10586 taught by Professor Jilinwang during the Fall '09 term at Zhejiang University.
- Fall '09
- Computer Science