Unified Graph Structured Models for Video Understanding
                 Anurag Arnab    Chen Sun Cordelia Schmid
                           Google Research
                   {aarnab, chensun, cordelias}@google.com

            Abstract                 sider the man who is speaking but no longer in the scene.
                                 And to correctly infer that the woman is “driving” the car,
  Accurate video understanding involves reasoning about     rather than “riding” like the man,  ...