Learning to Recognize Actions on Objects in Egocentric Video with Attention Dictionaries | Read Paper on Bytez