Human Activity Recognition: Fight Detection using Long term Recurrent Convolutional Network (LRCN)

Human Activity Recognition for Fight Detection is an important research domain aimed at automatically identifying patterns indicative of physical altercations. Leveraging deep learning models like CNNs and RNNs, this approach extracts features from video frames to recognize fight-related behavioural patterns. The primary objective is to develop a machine learning model capable of autonomously detecting instances of fights in video footage, using techniques such as the Long-term Recurrent Convolutional Network (LRCN). Training involves a comprehensive dataset encompassing examples of both fights and non-fight activities, with model performance evaluated using standard metrics. Each frame undergoes individual analysis by the model to predict the presence of any indications of a fight. The model predicts the action with an accuracy of 98.03% and the movements given as input can be categorized as fights or no fights.