Abstract

SHrack is a web-based service designed for fitness enthusiasts, providing real-time tracking and recording of home training exercises. By utilizing computer vision, SHrack enables users to independently track their fitness progress and ensure accurate count during repetitive exercises.

Introduction

In the realm of fitness, especially during weight training, it is often challenging to keep an accurate count of repetitions. While existing programs measure exercise counts, finding a service that offers real-time video streaming for accurate exercise count detection and management is rare. SHrack addresses this gap by combining computer vision with user-friendly web services.

Method

SHrack employs MobileNet and the Contextual Prediction Module (CPM) to extract heatmaps and Part Affinity Fields (PAF) based on 19 crucial body keypoints. Due to unsatisfactory results from pre-trained models, supervised training with a labeled dataset was performed, leading to fine-tuned posture estimation for accurate exercise count tracking.

Features & Implementation

Key features:

1. Real-time Posture Detection — Analyzes 19 crucial body keypoints to provide real-time feedback on exercise posture.
2. Exercise Count Tracking — Accurately counts repetitions so users can focus on exercise rather than counting.
3. User-friendly Interface — Web-based tool offering seamless access and use.

Conclusion

SHrack represents a step forward in the fusion of fitness and technology. By leveraging computer vision, it offers users an accurate exercise tracking experience that promotes better habits.