This project implements multi-object tracking by combining YOLOv8 for real-time object detection with DeepSORT for object tracking and re-identification. It runs the Ultralytics YOLOv8 model (via PyTorch) for detecting objects frame-by-frame, then uses DeepSORT’s deep appearance descriptor to maintain consistent object IDs across frames. The system supports tracking with visual trails, vehicle counting, and speed estimation on video streams.