Show Me What I Like: Detecting User-Specific Video Highlights Using Content-Based Multi-Head Attention | Read Paper on Bytez