Real-time violence detection using CNN and LSTM

There is a need for developing deep learning solutions to analyze videos to identify the presence of violence. In this paper, the authors propose a deep neural network for the recognition of violent videos. However, despite recent developments in deep learning, very few techniques based on deep learning have been proposed to address the problem of detecting violence from videos. The solution will play a major role in transforming the way law enforcement works and also supports the government’s initiative to make cities smarter. The model includes CNN and LSTM as a temporal relation learning method which allows us to capture localized spatiotemporal features which further makes it possible for us to analyze local motion taking place in the video. This work also focuses on accuracy and fast response time. The performance was evaluated on the hockey fight dataset as of now in terms of recognition accuracy.