124. PROBABILISTIC ACTIVE LEARNING-BASED DETECTION OF VEHICLES BY PARTS

Department: Electrical & Computer Engineering
Research Institute Affiliation: California Institute for Telecommunications and Information Technology (Calit2)
Faculty Advisor(s): Mohan Trivedi
Award(s): Honorable Mention

Primary Student
Name: Sayanan Vinoth Sivaraman
Email: ssivaram@ucsd.edu
Phone: 858-822-0002
Grad Year: 2013

Abstract
In this study, we introduce a novel probabilistic active learning approach to detection and tracking of vehicles by parts, with safety applications for intelligent vehicles. Vehicles are defined by a combination of parts, detected in real time using standard classifiers. A parts-based probabilistic representation allows for detection of vehicles by parts, and also allows for detection of occluded vehicles. Vehicles can be detected and tracked in a seamless framework, in both monocular and stereovision modalities, allowing for real-time 3D vehicle tracking in the on-road environment. Parts based modeling is robust to partial occlusions, and robustly identifies vehicles that are entering the sensors field of view. Experimental results show great promise for intelligent vehicles.

Related Links:

  1. http://cvrr.ucsd.edu/sayanan/

« Back to Posters or Search Results