Violence Detection Using Convolution Neural Network and Long Short-Term Memory

To detect aggression in video surveillance data, this research gives a novel methods that combines the two deep learning approaches that is Convolution Neural network (CNN) and Long Short-Term Memory (LSTM). The method successfully recognize violent actions by utilizing the advantages of Long Short Term Memory for temporal sequence learning and Convolution Neural Network for spatial features extraction. Pretty positive outcomes are obtained when the CNN and LSTM networks are combined for violence detection. Preprocessing includes normalization and frame extraction then a CNN network is used to extract and encode the spatial data from each frame. LSTM network is trained to represent the temporal connection between frames in order to distinguish between normal and abnormal sequences. The main purpose to combine CNN and LSTM is to create an accurate violence detection system. The suggested model outperforms the conventional techniques once it has been trained and verified using available dataset of 2000 videos. The results of the proposed model demonstrate the abilities for identifying violent from videos. The model has significant promise for enhancing security and safety standards through the use of state-of- the-art machine learning approach with 98.3% accuracy rate.