Camera with Python

A Python project using OpenCV and MediaPipe for real-time hand and face detection via webcam.

Features

• Hand Detection: Detection and tracking of up to 4 hands simultaneously
• Face Detection: Face detection with bounding boxes
• Gesture Recognition: Real-time gesture recognition using LSF (Langue des Signes Française)
• Real-time Interface: Live display via webcam

Project Files

• cam.py: Complete version with hand AND face detection
• camera.py: Simplified version with hand detection only
• signe.py: Gesture recognition using MediaPipe Gesture Recognizer
• predict.py: Sign language prediction using trained ML model
• hand_detection.py: Data collection tool for capturing hand landmarks
• train.py: Train the LSF classifier model
• download_models.py: Download required MediaPipe models

Installation

Prerequisites

• Python 3.7 or higher
• Connected webcam
• Internet connection (for downloading MediaPipe models)

Quick Start (All Platforms)

1. Clone or download this project

2. Run the installation script:

   Windows:

   install.bat

   Linux/macOS:

   chmod +x install.sh
   ./install.sh

   This will:

   • Create a Python virtual environment
   • Install all required dependencies
   • Download MediaPipe model files (hand_landmarker.task, gesture_recognizer.task)

Manual Setup

Create Virtual Environment

Windows:

python -m venv venv
venv\Scripts\activate

Linux/macOS:

python3 -m venv venv
source venv/bin/activate

Install Dependencies

pip install -r requirements.txt

Download MediaPipe Models

python download_models.py

This downloads:

• hand_landmarker.task - Hand landmark detection model
• gesture_recognizer.task - Gesture recognition model

Usage

Complete Version (hands + faces)

python cam.py

Simplified Version (hands only)

python camera.py

Gesture Recognition (LSF)

python signe.py

Sign Language Prediction

Requires a trained model (lsf_model.pkl). Train it first with:

python hand_detection.py
python train.py
python predict.py

Workflow: Training Custom Sign Language Model

1. Collect hand landmark data:

   python hand_detection.py

   • Press 's' to save hand positions
   • Press 'a' to quit
   • Data saved to data/ folder

2. Train the model:

   python train.py

   • Creates lsf_model.pkl
   • Displays training accuracy

3. Use the trained model:

   python predict.py

Controls

• Press 'a' to quit the application
• Press 's' (in hand_detection.py) to capture hand landmarks
• Camera window displays in real-time

Main Dependencies

• OpenCV: Image processing and video capture
• MediaPipe: Hand, face, and gesture detection
• NumPy: Numerical computations
• scikit-learn: Machine learning models

Technical Notes

• Default resolution: 640x480 pixels
• Hand detection confidence: 0.7
• Hand tracking confidence: 0.4
• Support for up to 4 hands simultaneously
• ML Model: Random Forest Classifier with 200 estimators

Troubleshooting

• "No module named mediapipe": Run pip install -r requirements.txt
• "hand_landmarker.task not found": Run python download_models.py
• Camera not working: Ensure webcam is connected and not in use by another app
• Slow performance: Reduce video resolution or decrease detection confidence