Human Body Tracking Method Based on YOLOV5s Object Detection

Autor: Shaymaa Tarkan Abdullah, Bashar Talib AL-Nuaimi, Hazim Noman Abed
Jazyk: angličtina
Rok vydání: 2023
Předmět:
Zdroj: Academic Science Journal, Vol 1, Iss 4 (2023)
Druh dokumentu: article
ISSN: 2958-4612
2959-5568
DOI: 10.24237/ASJ.01.04.692C
Popis: Body tracking is a viable solution for interacting with the human-computer and augmented reality. They are considered a necessity for a comprehensive understanding of human mobility. According to a robust tracking system of the human body, locating and tracing the human body in practical applications are challenging due to the enormous number of deformations and variations in body parts, postures, skin colures, lighting conditions, and clothes. To reduce complexity, several earlier works have concentrated on specific issues, such as face recognition, hand motion recognition, and mark recognition. The researchers agreed that the most previous approaches presume that people act while standing. Focusing on the detection and tracking of the human body, and the other component focusing on the detection and analysis of human action, the suggested system is based on the integration of tracking with the study of human activities through movement. And use YOLOV5s algorithm for detection and tracking , the result achieve by this algorithm and proposed system mAp 99%. That is replacing typical processing procedures like roboflow with an alternative that can function without the Internet. The proposed system is based on ten classes that contain a collection of overlapping verbs and can operate on photos, videos, and real-time systems.
Databáze: Directory of Open Access Journals