Spatial, Temporal and Casual Inference for Understanding Images and Video

Be the first to like this page
Images and Video, PROF SONG-CHUN ZHU, UCLA: Statistics – CS Department CS 201: Spatial, Temporal and Casual Inference for Understanding Images and Vid

Prof Song-Chun Zhu
UCLA Jointly: Statistics – Computer Science Department

In the race of machine vs. human, computers are gaining grounds from symbolic world like playing chess, to the real world like jeopardy and autonomous driving. The last highland humans hold perhaps will be vision, thank to the vast amount of visual knowledge we have and the inference power that a large portion of our brain is committed to. But this may not last very long. In this talk, I’d like to address two problems by reviewing our research in vision, cognition, and learning:
  • With literally unlimited amount of images and videos, can a computer learn the spatial, temporal and causal concepts and visual knowledge in an unsupervised way?
  • What are the learning principle and innate capacity, which are necessary and sufficient for a computer to jump start?

I will review our progress by showing demos on
  • deep spatial, temporal and causal parsing for understanding objects, scene and events
  • reasoning human intents
  • generating narrative text description (and speech) about the scene and events
  • answering human queries
Student Reviews (0)
  • Get Admissions Info On University of California - Los Angeles
  • What area of study interests you the most?

Date: May 10, 2012
Time: from 04:15 PM to 05:45 PM
Place: 400 Boelter Hall
Contact name:
Contact phone: 310 825-4033