Gait recognition is a biometric technique that identifies individuals based on their walking patterns. Unlike anothermethod such as face recognition, gait recognition can operate effectively even when facial features are not clearly visible, making it suitable for long-distance and surveillance applications. This project presents an advanced multi-person gait recognition system capable of detecting, tracking, and identifying individuals in real time. The proposed system integrates multiple technologies to achieve accurate recognition. First, human subjects are detected using a deep learning-based object detection model,followed by a tracking mechanism that assigns a unique identity to each individual across frames. Human pose landmarks are then extracted using a pose estimation framework to capture body movement dynamics. These features are processed through a deep learning model based on Long Short-Term Memory (LSTM) networks and attention mechanisms to learn temporal walking patterns.
