yo-WASSUP/Good-GYM

What it solves

Good-GYM is an AI-powered fitness assistant that automates the tracking of workout repetitions. It eliminates the need for manual counting during exercises like squats, push-ups, and sit-ups, providing users with real-time feedback on their form via visual skeleton overlays.

How it works

The system uses a standard webcam to capture video and employs the RTMPose model for high-accuracy, real-time pose detection. It tracks 17 key body points (based on the COCO format) to calculate joint angles. By comparing these angles against predefined thresholds in a JSON database, the app determines when a repetition is completed. The project is designed to run locally on the CPU for privacy and efficiency, though optional GPU acceleration is available for developers.

Who it’s for

It is designed for fitness enthusiasts who want to track their workout progress without manual counting and for developers who want a customizable pose-detection framework for fitness apps.

Highlights

• Real-time counting: Automatically tracks repetitions for a variety of exercises.
• Customizable exercises: Users can add new exercise types by editing a JSON configuration file without writing code.
• Privacy-focused: Runs entirely locally on the user's machine.
• Cross-platform: Available as a Windows portable executable, a Python source project, and an iOS application.