2014 Poster Sessions : A Unified Framework for Object Detection, Pose Estimation, and Sub-category Recognition

Student Name : Roozbeh Mottaghi
Advisor : Silvio Savarese
Research Areas: Artificial Intelligence
Abstract:
Object localization, 3D pose estimation, and sub-category recognition are three important problems in computer vision that are typically solved independently from each other. We propose a novel unified framework to tackle all these tasks simultaneously. There are advantages to jointly solving these correlated sub-tasks. For instance, knowing the viewpoint of an object prevents the detector from paying an unnecessary penalty for not detecting self-occluded parts, or knowing the sub-category of an object instance helps us localize the object more accurately since we can find a bounding box that better fits that particular sub-category. We show that our method is effective in performing these tasks jointly.

Bio:
Roozbeh Mottaghi is a postdoctoral researcher in the Computer Science department at Stanford University. He holds a PhD degree in computer science from UCLA. Prior to that, he obtained a Masters degree from Georgia Institute of Technology. His research interests are computer vision and machine learning, in particular, his focus is on applying statistical learning methods for object detection and pose estimation.