Camera Networks, Image Forensics and Reproducible Image Analysis: Emerging Challenges and Opportunities
Speaker: Professor B.S. Manjunath, University of California, Santa Barbara
Chaired by: Professor Mohan Kankanhalli, N-CRiPT Director
Date and Time: Wednesday 21 August, 2.00pm – 3.30pm
Venue: Seminar Room 7, COM1-02-07
The past few years have seen an AI revolution, mainly fueled by the resurgence of neural networks aided by significant advances in computing hardware. I will give an overview of our lab’s current work in computer vision that span a broad range of applications, from cybersecurity and media forensics to camera networks and activity recognition. The fundamental problems that cut across these applications are feature extraction and matching, and one can train neural networks to learn these effectively when large amounts of data is available. At the same time, these neural networks are quite brittle, and could be easily fooled. I will provide some recent results in understanding the stability of such networks. The presentation will conclude with an overview of BisQue, an open source scalable platform for scientific image analysis that was developed in our lab. BisQue evolved as part of our work microscopy imaging, but now is used in applications ranging from life science and medicine to marine sciences and materials science.
B. S. Manjunath received the Ph.D. degree in Electrical Engineering from the University of Southern California and is a Distinguished Professor of Electrical and Computer Engineering at the University of California, Santa Barbara and Director of the Center for Multimodal Big Data Science and Healthcare. His group has been at the forefront of scalable image analytics, having pioneered methods for image search and retrieval that have been incorporated into the ISO/MPEG7 standard. His group has contributed extensively to steganography, image and video segmentation, distributed camera networks, activity recognition, and most recently in image/video forensics and cybersecurity. He co-founded Mayachitra, Inc., a small business based in Santa Barbara that focuses on AI and computer vision. He has published over 300 articles that are highly cited in the literature and is a co-inventor on 24 patents. He has supervised 42 PhD theses at UCSB. He is a fellow of the IEEE and a fellow of the ACM.