Simple Bicep Curl Demo

A real-time bicep curl tracking application using MediaPipe pose estimation to count repetitions and provide visual feedback.

Features

• Real-time pose detection using MediaPipe
• Automatic rep counting for both left and right arms
• Visual feedback with colored landmarks
• SF Pro font rendering for clean, modern UI
• Dual-arm tracking with independent counting
• Configurable parameters for different exercise standards

Requirements

• Python 3.7+
• OpenCV
• MediaPipe
• NumPy
• PIL (Pillow)

Installation

1. Install the required dependencies:

pip install opencv-python mediapipe numpy pillow

2. Ensure you have the bicep_curl.py module in the same directory.

Usage

Run the demo:

python simple_bicep_curl_demo.py

Controls

• Press q to quit the application

Visual Indicators

The application displays:

• Rep counts for both arms in large, yellow text
• Colored landmarks:
  • 🔵 Blue circles: Shoulders
  • 🟢 Green circles: Elbows
  • 🔴 Red circles: Wrists
• Pose connections showing the full skeleton

Configuration

You can adjust these parameters at the top of the script:

IDEAL_ANGLE = 90      # degrees, for a good curl (arm bent at 90°)
TOLERANCE = 20        # degrees, allowed deviation
EXTENDED_THRESHOLD = 160  # degrees, arm considered straight

Parameter Explanation

• IDEAL_ANGLE: The target angle for a proper bicep curl (90° = arm bent at right angle)
• TOLERANCE: How much deviation from the ideal angle is acceptable
• EXTENDED_THRESHOLD: The angle at which the arm is considered fully extended for rep counting

How It Works

Rep Counting Logic

1. Start of rep: Arm must be extended (angle > 160°) and then begin curling
2. End of rep: Arm returns to extended position (angle > 160°)
3. Counting: Each complete cycle (extended → curled → extended) counts as one rep

Pose Detection

The application uses MediaPipe's pose estimation to track:

• Left shoulder (landmark 11)
• Left elbow (landmark 13)
• Left wrist (landmark 15)
• Right shoulder (landmark 12)
• Right elbow (landmark 14)
• Right wrist (landmark 16)

Angle Calculation

The elbow angle is calculated using the three-point method:

• Point A: Shoulder
• Point B: Elbow (vertex)
• Point C: Wrist

The angle at the elbow is computed using vector mathematics.

Technical Details

Dependencies

• MediaPipe: Google's framework for building perception pipelines
• OpenCV: Computer vision library for image processing
• NumPy: Numerical computing for angle calculations
• PIL: Python Imaging Library for font rendering

Font Support

The application attempts to load SF Pro font from macOS system locations:

• /System/Library/Fonts/SF-Pro-Text-Regular.otf
• /System/Library/Fonts/SF-Pro-Text-Bold.otf
• Falls back to system default fonts if SF Pro is unavailable

Performance

• Optimized for real-time processing
• Configurable detection confidence thresholds
• Efficient landmark tracking with MediaPipe

Troubleshooting

Common Issues

1. "No person detected": Ensure you're visible in the camera frame
2. Poor tracking: Adjust lighting conditions or camera position
3. Font errors: The app will fall back to default fonts if SF Pro is unavailable

Camera Setup

• Position yourself 2-3 feet from the camera
• Ensure good lighting
• Stand with your side profile visible for best tracking
• Keep both arms visible in the frame

File Structure

├── simple_bicep_curl_demo.py    # Main demo application
├── bicep_curl.py                # BicepCurl class for angle calculations
├── example curl tracking mediapipe.png  # Example screenshot
└── README_simple_bicep_curl_demo.md     # This documentation

Contributing

Feel free to modify the parameters or add new features:

• Adjust rep counting thresholds
• Add audio feedback
• Implement exercise form analysis
• Add data logging capabilities

License

This project is part of the AI Gym Pose Estimation suite for fitness tracking and analysis.