Advanced Video Surveillance with Temporal and Spatial Action Recognition for Harm Anticipation

Information Science & Engineering

In order to address the difficulties in differentiating between safe and dangerous behaviors in dynamic and complicated real-world contexts, this study suggests a unique framework for action recognition in films that uses deep learning techniques. The system integrates spatial and temporal data across several time scales by building upon the hierarchical structure of C3D convolutional neural networks. The model improves the early detection of risky activities by preprocessing video data and correcting potential biases. Applications such as surveillance, human-computer interaction, and sports analysis might benefit from the system's exceptional ability in identifying human activities, which has been confirmed using benchmark datasets (e.g., UCF101, HMDB51).

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.