Real-Time Workout Posture Correction using OpenCV and MediaPipe

Abstract

This paper proposes a program that uses OpenCV and MediaPipe to guide the workout (squat, push-up) posture correction content and the number of workouts by estimating the posture of real-time images. Specifically, the program estimates the landmarks of the user's body from the images acquired using the webcam in real time. Furthermore, it calculates the body angle and numerical values necessary to determine whether the posture is correct, and guides the user with the necessary corrections when the wrong posture is taken. It also increases and guides the number of workouts when the user takes the correct posture. Guidance for posture correction is displayed on the screen, and the number of workouts is guided by screen and voice. This program helps people exercise at home while checking their posture and work out with the proper posture without the help of an exercise trainer.