A Fitness Education and Scoring System Based on 3D Human Body Reconstruction

ABSTRACT

Affected by the Covid-19 epidemic, online fitness education has attracted a large number of users. However, when there are a large number of students in a same online classroom, it is difficult to get the coach’s advice and scores in time. To overcome this problem, we propose an AI fitness education system that uses 3D reconstruction technology to restore the shape of the human body and its bones. The skeleton is used for posture scoring. The 3D human body model is reconstructed by our improved VIBE network, with the accurate posture, shape and movement of the coach and students. By adding the loss function of the end limbs to the 3D human body model, compared with the performance of the original VIBE, we reduce the jitter noise in continuous motion. The training results show the accuracy of our improved VIBE. In addition, we have also established a scoring system, which can score the posture of the trainees based on the coach’s posture, and provide feedback through visual tag points. The experimental results show that our method is feasible and worthy of further exploration. © 2021, Springer Nature Switzerland AG.