Real Time Object Detection Tracking using YOLO and Deep SORT

Information science and engineering

This paper presents a comprehensive system for real-time object detection and tracking utilizing the YOLOv8 neural network architecture. The proposed system supports multiple input sources including static images, video files, realtime camera feeds, and phone-based streams. A Streamlit-based web interface provides intuitive user interaction for seamless integration across different application scenarios. The system incorporates adjustable confidence thresholds for flexible object detection tuning, and stores detection results in a structured database for historical tracking and analysis. Our implementation demonstrates efficient processing of multi-source inputs with real-time visualization and comprehensive object metadata collection including spatial dimensions and temporal information. The system achieves robust detection performance across various scenarios while maintaining user-friendly interface design. Experimental evaluation demonstrates 87-92% mAP accuracy with 6-15ms inference latency on GPU-equipped systems.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.