: Applying transformer architectures to video recognition.
The video filename is a specific clip from the Something-Something V2 dataset [1, 3]. This dataset is widely used in computer vision research to train models on human-object interactions and temporal reasoning [2, 4]. g60917.mp4
The video is used to help AI understand "visual common sense"—for example, knowing that an object will fall if pushed off an edge [2, 5]. Common Research Uses : Applying transformer architectures to video recognition
: Learning temporal aspects of video via self-attention. g60917.mp4