HCMMML_VideoCaptioner Links http://www.robots.ox.ac.uk/~vgg/research/very_deep/ Actions Smile/Grin Gaze/Glance/Watch Approach Frown/ model_id4.json/t7 is used as the base model