Automated Gym Exercise Form Checker: Deep Learning-Based Pose Estimation
摘要
The prevalence of weight training exercise as a fitness regimen has surged in popularity, encompassing individuals of diverse age groups, ranging from adolescents to seniors, all striving to maintain their physical well-being. While weight training exercises yield numerous health benefits, they can be counterproductive when executed improperly. Hence, an application “Automated Form Checker” is proposed to classify the exercise forms of users in the gym. The application utilizes video footage acquired from a closed-circuit television camera, facilitating object detection, and pose estimation on the video content. Subsequently, the derived sequential data are channeled into a multivariate long short-term memory model, which classifies users’ forms as correct or incorrect. The proposed application was able to deliver promising results with an accuracy of 98.88%.