Date: 2016-12-30 2:20pm-3:30pm

Location: R102, CSIE

Speaker:  Prof. Jia-Bin Huang

The recent success of deep learning in computer vision rely on massive amounts of manually labeled images. Natural videos contain significantly more information than images and are readily available. The additional time dimension, however, imposes new challenges such as increased space complexity and spatiotemporal appearance variations. In this talk, I will present how we can extract supervisory signals from unlabeled videos and show examples that leverage the scene dynamics for (1) synthesizing missing regions in videos in a temporally coherent manner, (2) tracking multiple faces in unconstrained videos, and (3) unsupervised learning of optical flow. I will end the talk by highlighting challenges and recent trends.

Biography

Jia-Bin Huang is an Assistant Professor in the Bradley Department of Electrical and Computer Engineering at Virgina Tech. He received his Ph.D. degree in the Department of Electrical and Computer Engineering at University of Illinois, Urbana-Champaign in 2016. His research interests span computer vision, computer graphics, and machine learning. Huang's research received the best student paper award in IAPR International Conference on Pattern Recognition in 2012 and the best paper award in the ACM Symposium on Eye Tracking Research & Applications in 2014.
Personal website: https://filebox.ece.vt.edu/~jbhuang