Workout Mentor: Improving posture in real-time using computer vision

摘要

Fitness workouts are tremendously helpful in improving individual health and fitness; yet, if practiced incorrectly by the user, they can be unproductive and potentially harmful. These blunders occur when the person fails to apply the correct posture or position. We propose “Workout mentor”, an application for maintaining proper posture while performing workout routines or yoga. Current work on workout monitoring depends heavily on personal trainers or wearables. Drawing inspiration from medical advancements like robotic surgery, we created an innovative application called “Workout mentor” that can direct workouts at the ease of your home using a single reference image. Workout mentor employs a machine learning approach to provide a real-time fitness instruction environment. Our research exploits the potential of posture estimation algorithms by providing real-time guidance for workout and yoga enthusiasts. Proposed model achieves 98.33% recognition accuracy. The model utilizes and learns from the crucial components that are necessary for all types of exercise.