Vu Hoang Minh Doan

Introduces AADC-Net, a multimodal deep neural network leveraging pretrained vision-language models, large language models, and object detection (DETR) for real-time anomaly detection and categorization in surveillance videos. The framework addresses data scarcity, imbalance, and computational challenges, demonstrating state-of-the-art performance on multiple datasets, with practical deployment in smart gyms and healthcare settings.