Local features for action recognition

Furqan M. Khan


Abstract

I'll be discussing use of local features for action recognition, specifically focusing on approach taken by Dr Laptev and his group at INRIA, France. The talk will cover definition of low-level features for spatiotemporal volumes and classification of actions using these features based on following three papers:
  • Learning realistic human actions from movies. Laptev et al. CVPR 2008
  • Retrieving actions in movies. Laptev and Perez, ICCV 2007
  • On Space-Time Interest Points. Laptev, IJCV 2005


    Maintained by Qian Yu