Optimization of a web-based real-time and file-based human detection system Using YOLOv8 and flask framework

Siahaan, Andysah Putera Utama and Waruwu, Rido Favorit Saronitehe and Putra, Heri Eko Rahmadi (2025) Optimization of a web-based real-time and file-based human detection system Using YOLOv8 and flask framework. Global Journal of Engineering and Technology Advances, 24 (1). pp. 144-150. ISSN 2582-5003

This study discusses the development and optimization of an object detection system using the YOLOv8 (You Only Look Once version 8) algorithm integrated with the Flask framework in a web-based environment. The system supports two main modes of operation, namely image upload and direct detection using the camera. The main goal of the system is to provide accurate, efficient, and accessible solutions without the need for local installation. The test was conducted using 50 test images with different variations in background, lighting conditions, and the number of human objects, resulting in an average inference time of 0.43 seconds per image, precision of 95.1%, recall of 91.7%, and mAP@0.5 of 93.4%. In real-time testing, the system was able to run stably with a video processing speed of 18-22 frames per second. These results show that the developed system has high performance and is feasible to apply for online visual monitoring needs. Potential system development includes the addition of object tracking features, automatic log storage integration, and real-time notification systems to expand usability in various fields.