Researchers have developed a new two-stage framework for real-time video anomaly detection that utilizes YOLO Pose Estimation and CLIP-based semantic scoring. This method achieves a throughput of approximately 51 FPS on an NVIDIA Titan XP GPU, offering a significant speedup over existing baselines. The system demonstrates strong performance on various datasets, maintaining high AUROC values while eliminating the need for optical flow or density-based scoring modules. AI
IMPACT This framework could improve real-time security and monitoring systems by enabling faster and more accurate anomaly detection.
RANK_REASON The cluster contains a research paper detailing a new AI model and framework. [lever_c_demoted from research: ic=1 ai=1.0]
- Chulalongkorn University
- CLIP ViT-B/32
- CUHK Avenue
- ShanghaiTech Campus
- Vanodhya G. Warnasooriya
- YOLO v11n-pose
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →