Application of YOLOv4 object detection models in biomechanics problems of human kinematics

In this work, the problem of automating the analysis of the kinematics of the human body has been solved. Data collection was carried out in the form of video recording. The classification of areas of the human body was performed using convolutional neural networks of three different architectures. The developed system solved the classification of target zones of the human body with a precision 0.98 and recall 0.97. This result indicates the ability of the system to analyze the kinematics of the human body.