1 paper · 1 filter
Keliang Li, Yansong Li, Hongze Shen +3
The dense, temporal nature of video presents a profound challenge for automated analysis. Despite the use of powerful Vision-Language Models, prevailing methods for video understan…