MSR VTT 2016

AbstractWe present a video captioning approach that encodes features by progressively completing syntactic structure (LSTM-CSS). To construct basic syntactic structure (i.e., subject, predicate, and object), we use a Conditional Random …...