Computer-Aided Assessment System for Calisthenics Exercises

Abstract

Amidst the global pandemic, Physical Education classes had to resort to online submission of exercise performance videos, which is a more tedious and time-consuming task for PE instructors to individually view and assess. As a possible tool to assist with this problem, this paper presents a computer-aided assessment system that outputs a summarized assessment for three calisthenics exercises, namely the push-up, the squat, and the sit-up. The system uses OpenCV as a platform for computer vision and utilizes Mediapipe's pose detection library to detect the pose of the target within a video. By calculating the angle between the landmark body parts associated with the execution of a specific exercise, the system can count the reps as well as provide insight on the angle depth of the exercise execution and its range of motion. It outputs this data as an Excel spreadsheet for convenience since the system is intended to assist Physical Education instructors with checking the exercise videos of their students. The system was validated by comparing its summary output with that of a human PE professor and statistically treated using SPSS. The results show that it provides a similar rating to that of the human in terms of counting total number of reps and correct reps for push-ups and squats but had some issues with assessing range of motion for sit-ups.