Arms Detector

An interactive computer vision application that detects arm motion using your webcam and displays a floating video panel when movement is detected.

Features

• Real-time arm motion detection using MediaPipe Pose landmarks
• Fallback motion detection via OpenCV when MediaPipe is unavailable
• Floating video panel that appears when arm movement is detected
• Cross-platform camera support with Windows DirectShow/MediaFoundation backends
• Adjustable sensitivity on the fly with keyboard controls
• Visual HUD showing motion levels and detection status
• Skeleton overlay for visual feedback on tracking

Demo

When you move your arms, a floating video panel appears in the corner of the screen. The panel hides when you stop moving.

Installation

Requirements

• Python 3.8 or higher
• Webcam

Dependencies

pip install opencv-python mediapipe

Usage

Basic Usage

python main.py --video myclip.mp4

Command Line Options

python main.py --video <path> [--sensitivity <value>] [--camera <index>]

Option	Description	Default
--video	Path to the video file to display	Required
--sensitivity	Motion threshold (lower = more sensitive)	0.04
--camera	Camera device index	0

Examples

# Use default camera with a video file
python main.py --video tutorial.mp4

# Use a specific camera and higher sensitivity
python main.py --video workout.mp4 --camera 1 --sensitivity 0.02

# Less sensitive (requires more movement to trigger)
python main.py --video demo.mp4 --sensitivity 0.08

Controls

Key	Action
Q or ESC	Quit the application
+	Increase sensitivity (lower threshold)
-	Decrease sensitivity (higher threshold)

How It Works

1. Camera Capture: Opens your webcam with optimized settings for real-time processing
2. Motion Detection:
   • Uses MediaPipe Pose to track wrist positions relative to shoulders
   • Falls back to OpenCV motion detection in side bands if MediaPipe unavailable
3. Smart Triggering:
   • Calculates torso-relative wrist movement to ignore body sway
   • Applies smoothing over multiple frames to reduce false positives
4. Video Overlay: Displays a floating "Now Playing" panel when motion exceeds threshold

Project Structure

hands-detecter/
├── main.py          # Main application code
├── myclip.mp4       # Sample video file
└── README.md        # This file

Troubleshooting

Camera not detected

• Check Windows camera privacy settings
• Try a different camera index: --camera 1
• Ensure no other application is using the camera

Black frames

• The app will automatically try different camera backends
• Check if your camera works in other applications
• Try adjusting room lighting

High CPU usage

• Lower the camera resolution in main.py (change PANEL_W, PANEL_H)
• Use a lower frame rate video file

Technical Details

• Primary tracking: MediaPipe Pose landmarks (wrists, shoulders, nose)
• Fallback tracking: OpenCV optical flow in side ROI bands
• Smoothing: 6-frame rolling average for wrist positions
• Threading: Separate threads for camera and video file reading
• Platforms: Windows (DirectShow/MediaFoundation), Linux, macOS

License

MIT License - feel free to use and modify as needed.

Author

Created by GitMehdi-sys