PPE Compliance Monitoring System

Project Overview

This system detects Personal Protective Equipment (PPE) compliance using computer vision (YOLOv8) and provides a web interface for monitoring violations, viewing statistics, and managing alerts. It includes real-time notifications via Telegram.

This project was refactored to meet industry standards, improve reliability, and provide a professional user interface suitable for demonstration purposes.

Features

• Media Upload: Upload images (JPG, PNG) or videos (MP4, AVI, MOV, WebM) via a web interface.
• PPE Detection: Utilizes a fine-tuned YOLOv8 model to detect various PPE items and identify violations (e.g., missing hard hats, vests, masks based on configured rules).
• Violation Logging: Records detected violations in an SQLite database, including timestamp, violation type, location, severity, and a snapshot image.
• Web Dashboard: Displays key statistics about violations:
  • Counts by equipment type
  • Counts by severity (High, Medium, Low)
  • Counts by location
  • Counts by status (Unresolved, Investigating, Resolved)
  • Daily violation trend chart (last 30 days)
• Violation Log: A filterable/sortable table view of all recorded violations with links to evidence images and status update capability.
• Telegram Notifications: Sends real-time alerts to a configured Telegram chat when violations are detected (includes violation details and image). Features a cooldown mechanism to prevent notification spam.
• Configurable: Easily configure model paths, database location, Telegram credentials, PPE class mappings, violation rules, and area requirements via a .env file.

Setup and Installation

1. Clone the repository:

   git clone <repository-url>
   cd ppe-detection-project

2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install dependencies:

   pip install -r requirements.txt

   Note: ultralytics and opencv-python-headless can be large.

4. Configure Environment Variables:

   • Copy .env.example to .env: cp .env.example .env
   • Edit .env and fill in the required values:
     • SECRET_KEY: Generate a strong secret key (python -c 'import os; print(os.urandom(24).hex())').
     • TELEGRAM_BOT_TOKEN: Your Telegram Bot token.
     • TELEGRAM_CHAT_ID: The chat ID where notifications should be sent.
     • MODEL_PATH: Verify the path to your trained .pt file.
     • PPE_CLASS_MAPPING: Crucially, update this JSON string to match the class IDs and names from the data.yaml file used to train your specific YOLOv8 model.
     • VIOLATION_CLASSES: Update this JSON list with the exact class names (from the mapping above) that represent a violation (e.g., "NO-Hardhat").
     • AREA_REQUIREMENTS: Define required PPE for different work areas if needed.
     • Review other paths if you changed the structure.

5. Place the Trained Model:

   • Ensure your working trained YOLOv8 model file (e.g., last.pt or best.pt) is placed in the models/ directory and renamed to match the MODEL_PATH in your .env file (e.g., models/yolov8s_ppe_custom.pt).

6. Initialize the Database:

   • The database (violations.db) and schema will be created automatically when you first run the application using run.py.

Running the Application

1. Activate the virtual environment:

   source venv/bin/activate

2. Run the Flask development server:

   python run.py

3. Access the application: Open your web browser and navigate to http://127.0.0.1:5000 (or http://<your-server-ip>:5000).

Usage

1. Navigate to the Upload page.
2. Enter the Location and select the Area Type.
3. Choose an image or video file and click Upload and Analyze.
4. Wait for processing. You will be redirected to the Violations Log.
5. Check the Dashboard for statistics and charts.
6. Monitor your configured Telegram chat for real-time violation notifications.
7. On the Violations Log page, you can view violation details, see the evidence image, and update the status (Unresolved, Investigating, Resolved).