MG-TVMF: Multi-grained text-video matching and fusing for weakly supervised video anomaly detection

Key Points

• Video anomaly detection improves with multi-grained text-video matching and fusing techniques.
• Utilizing weakly supervised learning, the method enhances detection accuracy, indicating effectiveness in diverse scenarios.
• The analysis demonstrates that integrating multi-grained features can achieve better results than traditional methods.
• This approach may enable more reliable and efficient monitoring in real-time video applications.