HandSign-Detection

Real-time hand sign language detection using OpenCV, MediaPipe, and deep learning. Detects American Sign Language letters I, L, O, V, E, U from webcam input in real-time.

Features

• Real-time detection - Live webcam hand sign recognition
• Data collection - Capture custom hand sign training data
• Deep learning model - Keras/TensorFlow classification model
• Preprocessing - Automatic hand cropping and normalization

Tech Stack

• Python
• OpenCV
• MediaPipe (via cvzone)
• TensorFlow/Keras
• NumPy

Project Structure

HandSign-Detection/
├── Data Collection.py    # Script to collect hand sign training data
├── Testing.py            # Real-time hand sign detection & classification
├── Data/                 # Training data organized by class (I, L, O, V, E, U)
├── Model/
│   ├── keras_model.h5    # Pre-trained Keras model
│   └── labels.txt        # Class labels
└── .idea/                # IDE config

Quick Start

# Install dependencies
pip install opencv-python cvzone numpy tensorflow

# Collect training data (press 's' to save frames)
python "Data Collection.py"

# Run real-time detection
python Testing.py

How It Works

1. Data Collection - Uses MediaPipe to detect hands via webcam, crops the hand region, resizes to 300x300, and saves labeled images
2. Training - Model trained on collected data using Keras (saved as keras_model.h5)
3. Testing - Real-time inference: detects hand, preprocesses frame, classifies with trained model, and overlays prediction on screen

Labels

Index	Sign
0	I
1	L
2	O
3	V
4	E
5	U

License

MIT