Human Behaviour and Abnormality Detection using YOLO and Conv2D Net

In many fields, such as security, medical, and surveillance, human behavior and anomaly detection is essential. This abstract describes a novel method for reliably detecting and classifying anomalous human behavior that combines two deep learning algorithms: CONV2d net and YOLO (You Only Look Once). Modern object detection algorithms like YOLO are renowned for their great accuracy and real-time performance. A popular convolutional neural network architecture for image recognition applications is the CONV2d net. The major goal is to improve anomaly detection and human behavior accuracy and efficiency by combining these two methods.The suggested approach locates and effectively detects humans in real-time video feeds by using YOLO. Then, human behavior is classified using the CONV2d net into specified categories including standing, walking, running, and abnormal actions. The integration of these two methods enables reliable and precise identification of human behavior across a range of contexts. In addition, the system makes use of an extensive collection of annotated movies, which allows the deep learning models to be trained and validated. Through a series of comprehensive experiments, YOLO and CONV2d net fusion model performances in identifying and classifying human behavior is presented, including anomalous behaviors that might point to possible threats or dangerous situations.