Junwen Xiong

A new approach to identifying fake videos by combining audio, visual, and motion data.

A new method combines video saliency prediction and detection, improving performance.

DiffSal improves how computers predict viewer interest in videos.