Theft Guard AI is a comprehensive security solution designed for retail environments. It leverages Computer Vision and Artificial Intelligence to detect suspicious behaviors such as shoplifting, fighting, and item concealment in real-time. The system processes video feeds from multiple CCTV cameras and provides instant alerts to security personnel via a modern web dashboard, Email, and Telegram.
The core of the system is built on a multi-stage AI pipeline that runs locally:
- Object Detection: Utilizes optimized YOLOv8 models to identify person and retail objects. It can be extended with specialized models ('shoplifting.pt') to directly detect theft actions.
- Behavior Analysis: Tracks the movement of hands relative to pockets/bags to detect concealment attempts (Hand-to-Pocket gestures).
- Pose Estimation: Analyzes human skeleton points to detect suspicious postures, such as bending down in aisles or reaching into restricted areas.
- Blacklist Monitoring: Instantly identifies known offenders entered into the system database and triggers high-priority alerts.
- VIP Detection: Can be configured to recognize loyal customers or VIPs for personalized service.
built with Next.js, the dashboard offers a centralized control room experience:
- Live Video Feeds: Stream vertically or horizontally from multiple camera sources (USB or IP Cameras) simultaneously via WebSockets.
- Dynamic Graphs: Visualizes weekly and daily alert statistics to track security trends over time.
- Visual Alerts: Flashing on-screen notifications highlight the camera and timestamp of any detected event.
- Event Logging: Every detection is saved to a local SQLite database with a timestamp, alert type, and confidence score.
- Snapshot Capture: High-resolution images of the event are automatically saved for evidence.
- History Viewer: Browse past alerts, view snapshots, and export data directly from the interface.
Stay informed even when away from the desk:
- Telegram Integration: Sends an instant photo and caption to a specified Telegram Chat ID via Bot API.
- Email Reports: Dispatches detailed text alerts to configured email addresses using SMTP.
- Region of Interest (ROI): Users can draw custom polygons on the camera feed to define sensitive areas (e.g., cash registers, high-value shelves).
- Zone-Specific Rules: Detection sensitivity can be adjusted based on whether a person is inside or outside these zones.
- Backend: Python, FastAPI, OpenCV, Ultralytics YOLOv8, Face Recognition, Albumentations
- Frontend: Next.js 14, React, Tailwind CSS, Recharts, Lucide React
- Database: SQLite (Lightweight, local storage for events and faces)
- Communication: WebSockets (Real-time data), SMTP (Email), HTTPS (Telegram API)
- Python 3.9 or higher
- Node.js (LTS version)
- NVIDIA GPU with CUDA (Recommended for real-time performance)
Clone the repository and install the required Python packages:
git clone https://github.com/Start-Up-Vahap/Theft-Detection.git
cd Theft-Detection
pip install -r requirements.txtIf you have a specialized model (shoplifting.pt), place it in the root directory. Otherwise, the system defaults to the standard YOLOv8 model with behavior logic.
Navigate to the dashboard directory and install dependencies:
cd dashboard
npm installSimply run the helper script to launch both services:
start_system.batStart Backend API:
py backend.pyStart Frontend UI:
cd dashboard
npm run devOpen your browser and navigate to http://localhost:3000.
- Fork this repository.
- Create a feature branch (
git checkout -b feature/NewFeature). - Commit your changes (
git commit -m 'Add new feature'). - Push to the branch (
git push origin feature/NewFeature). - Open a Pull Request.
Distributed under the MIT License. See LICENSE for more information.
Developer: Abdulvahap Öğüt