Aplicação Web para detecção de objetos e rastreamento de trajetórias automatizada utilizando YOLOv8

Abstract

The growing demand for automated systems capable of performing accurate analyses on videos and images drives the development of advanced computer vision solutions. In this context, there is a need for efficient tools for object detection and tracking that can be applied in various fields such as security, transportation, industrial automation, and more. This study presents the development of a web application for object detection and trac king in videos and images using the YOLOv8 algorithm. The application processes files provided by users, applying computer vision techniques to efficiently identify and track objects. The application’s interface allows users to select from different YOLO models (nano, small, medium, large, and huge), each optimized for varying levels of precision and speed. Additionally, the application uses 12 of the 80 classes available in YOLO. These classes were prioritized based on their relevance in real-world scenarios where the detection and tracking of people, vehicles, animals, and common objects are essential. Comparative tests were conducted using the different YOLO models and the tracking algorithms BoT-SORT and ByteTrack to evaluate performance in terms of accuracy and processing time. The results include a processed file with the identification of detected objects and detailed data on the number of objects, detected classes, average detection ac curacy, and processing times for each configuration tested. The conclusion highlights the effectiveness of the developed system and proposes future improvements, such as model updates, integration with external links for real-time video processing, and enhancements to the user interface to facilitate result visualization.