You are here

Social Network Analysis

Despite countless efforts on video understanding, most efforts do not go beyond analyzing or grouping trajectories, or understanding individual actions performed by tracked objects. Generally speaking researchers have not considered video analysis from a sociological perspective.

Our research marks to first studies on social network analysis from video by learning the relations among actors in a video. We use visual and auditory information to quantify a grouping cue at the scene level, which serves as soft constraints among the actors. These soft constraints are then integrated to learn inter-actor affinity. The commu- nities in the resulting social network are discovered by subjecting the inter-actor affinity matrix to a generalized modularity principle.

We have also used the same concepts for analysis of transportation networks and sporting events.

Published papers

  1. L. Ding and A. Yilmaz. 2014. Learning Social Relations from Videos: Features, Models and Analytics.In Human-Centered Social Media Analytics. Edited by Y.R. Fu and S. Rees. New York, NY: Springer Verlag
  2. L. Ding and A. Yilmaz. 2011. Inferring Social Relations from Visual Concepts. IEEE International Conference on Computer Vision (ICCV). Barcelona, Spain. (November): 1-6.
  3. L. Ding and A. Yilmaz. 2010. Learning Relations Among Movie Characters: A Social Network Perspective. In: Proc. of European Conference on Computer Vision (ECCV). Crete, Greece. (September): 1-10

  4. K. Park and A. Yilmaz. 2010. Social Network Approach to Analysis of Soccer Game. In: International Conf. on Pattern Recognition (ICPR), oral presentation. Istanbul, Turkey: IAPR. (August 21)

  5. K. Park and A. Yilmaz. 2010. A Social Network Analysis Approach to Analyze Road Networks. In: ASPRS Annual Conference. San Diego, CA. (April 25)