Research

I am interested in the general problem of image understanding, which encompasses detection and recognition of objects, visual categories, and scenes. Image understanding requires successful integration of low, mid, and high level visual processing. Effectively combining cues from all levels is an open research problem that I hope to address.

My focus is on methods to exploit the web as a vision dataset. The recent explosion of easily accessible images from the internet begets new possibilities for vision research. Web images and video are labeled to varying degrees and accuracies. I am investigating ways to combine noisy labels and image features for automated dataset collection, search reranking, and image classification.

Recent Activities


Nicholas W. Morsillo

URCS grad student

Office:   CSB 627
Phone:   (585) 275-0922
Email:
Nick Morsillo, Chris Pal, and Randal Nelson. "Mining the Web for Visual Concepts." URCS Tech Report. [pdf]

SUnS: Scene Understanding Symposium 2008. Presented "Semi-Supervised Visual Scene and Object Analysis from Web Images and Text." [poster pdf]


Vision Seminar Talks

2/8/08 Study of "Learning Probabilistic Models for Contour Completion in Natural Images"
(X. Ren, C. Fowlkes, J. Malik - IJCV2007)
10/19/07 Study of "City-Scale Location Recognition"
(G. Schindler, M. Brown, R. Szeliski - CVPR2007)
9/21/07 Study of "TextonBoost for Image Understanding"
(J. Shotton, J. Winn, C. Rother, A. Criminisi - IJCV2007)
4/13/07 Study of "Sharing Features: Efficient Boosting Procedures for Multiclass Object Detection"
(A. Torralba, K. Murphy, W. Freeman)
2/23/07 Study of "Robust Object Recognition with Cortex-like Mechanisms"
(T. Serre, L. Wolf, S. Bileschi, M. Riesenhuber, T. Poggio)

TA Course Pages

Spring 2007 290A: Machines and Consciousness
Fall 2006 240: The computational brain

Links

URCS Computer Vision Seminar
rocwiki
google
last.fm
pandora





UPHOME
URCS People | URCS Home Page