Form Check: Exercise Posture Correction Application

Abstract

Weightlifting is a popular sport and a hobby of millions; however, the risk of injury is high, primarily when performed with incorrect form/posture due to voluntary or involuntary reasons. It results in a significant loss of strength and hinders progress. The gradual strain on tendons and ligaments causes wear and tear. This paper aims to develop a deep learning model that detects incorrect posture and provides insight into remedial steps that need to be taken to correct the same, thereby providing increased safety and efficiency. Our Form Check model detects a user's form, vector geometry of posture is evaluated and corrected. A graphical representation of the exercise is provided as feedback to the user. Object detection, pose estimation, and action recognition algorithms infer the human skeleton and recognize the exercise. The inconsistency in the exercise performed is determined and mapped to specific imperfections in the stance of users during the exercise through Graph