A paper entitled "Discriminative models for multi-class object layout" by PhD student Chaitanya Desai and Assistant Professors Deva Ramanan and Charless Fowlkes received the Marr Prize at the International Conference on Computer Vision (ICCV) held the first week of October in Kyoto, Japan.
The prize is awarded to the best paper at ICCV and is considered one of the top honors in computer vision. The award is named after David Marr, a theoretical neuroscientist who made profound contributions to the theory of both human and machine vision in the 1970's.
The paper describes research on a new approach to modeling contextual relations between objects in an image (e.g. bottles are often seen resting on top of tables but not the other way around). The system automatcially learns these relations from example images and uses this information to outperform existing approaches to object detection.
Fowlkes' research is in computational vision, both in understanding the information processing capabilities of the human visual system and in developing machine vision systems. He is also interested in applying computer vision techniques to automating the analysis of biological data and developing algorithmic tools for understanding morphology and spatial aspects of gene expression.
Ramanan's research interests span computer vision, machine learning, and computer graphics. His past work focused on the analysis of human movement from video, including tracking people and recognizing their actions. Current interests include object recognition, large-scale image/video processing, structured-prediction approaches to learning, and activity recognition.