A deep learning project to identify the violence from the videos (not real time).
We live in the society that relies heavily on the usage of the CCTV cameras for ensuring high security. However, such an approach is highly controversial, as we usually use CCTV footages, only hours or even days after the incident has occurred. These CCTV footages are rarely used to prevent crime or react to it in real-time. The reason for such an inefficiency is that the lack of human support for monitoring the huge amount of CCTV footages. An efficient surveillance system can be developed with the help of the technologies like computer vision, video-image processing and deep learning. Meanwhile the action recognition community has been focused mostly only on detecting simple actions like clapping, jogging, walking. The detection of aggressive behaviours is still a barrier because of its complex nature. Many real time problems can be solved by using machine learning and the data analytic techniques.
- Python 3.0
- tensorflow 2.0
- keras 2.3.0
- matplotlib
- ffmpeg