multi-head attention

Combining Global and Local Attention with Positional Encoding for Video Summarization

Α supervised self-attention-based method for automated video summarization.