People Counter Web App

A real-time web application that uses USB camera face detection to count people entering and leaving a room, with zone-based entry/exit tracking.

Features

• Real-time face detection using OpenCV DNN face detector
• Zone-based entry/exit tracking (left zone = entry, right zone = exit)
• Live video streaming via MJPEG
• Real-time count updates via WebSocket
• Visual indicators for overpopulation warnings
• Clean, modern web interface

Prerequisites

• Python 3.7 or higher
• USB camera connected to your computer
• Linux/Windows/macOS

Installation

1. Clone or download this repository

2. Install Python dependencies:

   pip install -r requirements.txt
   

3. Download the face detection model files:

   python download_models.py
   

   This will download the required OpenCV DNN model files to the models/ directory.

   Note: If the automatic download fails, you can manually download:

   • deploy.prototxt from: https://github.com/opencv/opencv/blob/master/samples/dnn/face_detector/deploy.prototxt
   • res10_300x300_ssd_iter_140000.caffemodel from: https://github.com/opencv/opencv_3rdparty/dnn_samples_face_detector_20170830/res10_300x300_ssd_iter_140000.caffemodel

   Place both files in the models/ directory.

Usage

1. Make sure your USB camera is connected

2. Run the Flask application:

   python app.py
   

3. Open your web browser and navigate to:

   http://localhost:5000
   

4. The application will:

   • Display live video feed from your camera
   • Detect faces in real-time
   • Count people entering (left zone) and exiting (right zone)
   • Display current occupancy and statistics
   • Show visual warnings when occupancy is high

Configuration

You can modify the following settings in app.py and camera.py:

• Camera index: Change camera_index=0 to use a different camera
• Frame processing rate: Adjust process_every_n_frames (default: 3) to balance performance and accuracy
• Face detection confidence: Modify face_confidence threshold (default: 0.5)
• Zone boundaries: Adjust entry_zone_percent and exit_zone_percent in zone_tracker.py
• Cooldown period: Change cooldown_seconds to prevent double-counting (default: 2.0 seconds)
• Maximum occupancy: Update MAX_OCCUPANCY in static/js/main.js for overpopulation warnings

How It Works

1. Camera Capture: The camera module captures frames from your USB camera
2. Face Detection: OpenCV DNN detects faces in the video frames
3. Zone Tracking: The zone tracker determines which zone each face is in:
   • Left 40% = Entry zone (green)
   • Right 40% = Exit zone (red)
   • Center 10% = Buffer zone (ignored to prevent false counts)
4. Counting Logic: People are counted when they appear in entry/exit zones with a cooldown period to prevent double-counting
5. Real-time Updates: Counts are sent to the web interface via WebSocket for live updates

Project Structure

PeopleCounter/
├── app.py                 # Flask main application
├── camera.py              # Camera capture wrapper
├── face_detector.py       # Face detection module
├── zone_tracker.py        # Zone-based tracking logic
├── download_models.py     # Script to download model files
├── requirements.txt       # Python dependencies
├── models/                # OpenCV DNN model files
│   ├── deploy.prototxt
│   └── res10_300x300_ssd_iter_140000.caffemodel
├── templates/
│   └── index.html         # Main web interface
└── static/
    ├── css/
    │   └── style.css      # Styling
    └── js/
        └── main.js        # Client-side JavaScript

Troubleshooting

• Camera not found: Make sure your camera is connected and try changing the camera_index in camera.py
• Model files missing: Run python download_models.py to download required files
• Slow performance: Increase process_every_n_frames value or reduce video resolution
• Double counting: Increase the cooldown_seconds value in zone_tracker.py

Security & Privacy

• All processing runs locally (no cloud storage)
• No face recognition - only detection (no personal identification)
• Video frames processed in memory, not stored
• Optional: You can modify the code to blur faces for display if privacy is a concern

License

This project is open source and available for use and modification.