ABNORMAL BEHAVIOUR DETECTION

In the realm of public safety and mental health, the recognition of abnormal or violent behaviour inindividuals with mental disorders within public spaces is a critical concern. The conventionalmethods utilizing Graph Convolutional Network (GCN) and 3D Convolutional Neural Network(3DCNN) algorithms have been instrumental, but their limitations impede their efficacy in dealingwith the nuanced complexities of real-world scenarios. Our proposed system introduces TransferLearning, leveraging pre-trained models to adapt efficiently to limited data, crucial in abnormalbehaviour detection. Combining this with the YOLO V10 algorithm revolutionizes object detection,processing entire images in one go for enhanced speed and accuracy. This integration significantlyimproves real-time abnormal behaviour detection in public spaces, crucial for the safety ofindividuals with mental disorders and those around them. By mitigating potential harm throughprecise detection, our system contributes to robust public safety solutions. Its implementation marksa milestone in addressing complex challenges associated with abnormal behaviour, promising moreeffective and adaptive solutions for public safety