Work

Efficient Multi-Head Attention for Human-Object Interaction Recognition and Video Scene Graph Generation