Mike Zheng Shou

Introducing AssistPDA, a pioneering framework for real-time online video anomaly prediction, detection, and analysis leveraging vision-language models with a novel spatiotemporal relation distillation module and constructed benchmark dataset VAPDA-127K.