Action Recognition In Still Images And Inference Of Object Affordances