Vietnamese License Plate Detection and Recognition System

A comprehensive YOLO-based system for detecting and reading Vietnamese license plates from images and live camera feeds. This project combines two deep learning models: one for plate detection and another for character recognition.

🚀 Features

• Real-time License Plate Detection: Uses YOLOv8 to detect license plates in live camera feeds
• Character Recognition: Reads and extracts text from detected license plates
• Automatic Cropping: Saves high-confidence detections for further analysis
• GPU/CPU Support: Optimized for both GPU and CPU inference
• Batch Processing: Process multiple images at once
• Confidence Thresholding: Configurable confidence levels for detection and cropping

📁 Project Structure

├── README.md                    # This file
├── TRAINING_SUMMARY.md         # Training progress and results
├── requirements.txt            # Python dependencies
│
├── Training Scripts/
│   ├── plate_train.py          # Train the license plate detection model
│   └── reader_train.py         # Train the character recognition model
│
├── Detection & Recognition/
│   ├── cropped_camera_detector.py  # Real-time camera detection with cropping
│   ├── gpu_detector.py             # GPU-optimized batch detection
│   ├── predict_plates.py           # Standalone plate prediction
│   └── reader_plates.py            # Character recognition from cropped plates
│
├── Models/
│   ├── yolo11n.pt              # Pre-trained YOLO11 base model
│   ├── yolov8n.pt              # Pre-trained YOLOv8 base model
│   └── yolo_training/          # Trained model weights
│       ├── vnlp_model7/        # Vietnamese license plate detection model
│       ├── LPR_model1/         # License plate reader model v1
│       └── LPR_model2/         # License plate reader model v2
│
├── Datasets/
│   ├── VNLP yolov8/           # Vietnamese license plate dataset
│   └── LP reader yolov8/       # Character recognition dataset
│
├── Output/
│   ├── detected_plates/        # Cropped license plates (high confidence)
│   ├── read_detected_plates/   # Processed plates with recognized text
│   ├── batch_predictions/      # Batch processing results
│   └── runs/                   # Training logs and metrics
│
└── Resources/
    ├── Photos/                 # Test images
    └── Videos/                 # Test videos

🔧 Installation

Prerequisites

• Python 3.8+
• CUDA (optional, for GPU acceleration)

Setup

1. Clone this repository:

git clone <repository-url>
cd Prototype4-Yolo

2. Install dependencies:

pip install ultralytics opencv-python torch torchvision numpy pillow

3. Ensure you have the trained models in the yolo_training/ directory

🎯 Usage

1. Real-time License Plate Detection

Detect and crop license plates from live camera feed:

python cropped_camera_detector.py

• ESC: Exit the application
• SPACE: Pause/Resume detection
• Automatically saves high-confidence detections to detected_plates/

2. Character Recognition

Read text from detected license plates:

python reader_plates.py

• Processes all images in detected_plates/
• Outputs annotated images to read_detected_plates/

3. Batch Processing (GPU)

Process multiple images efficiently:

python gpu_detector.py

4. Single Image Prediction

Test the model on a single image:

python predict_plates.py

🏋️ Training

Train License Plate Detection Model

python plate_train.py

• Uses Vietnamese license plate dataset (VNLP yolov8/)
• Trains YOLOv8 to detect license plate regions
• Saves model to yolo_training/vnlp_model*/weights/best.pt

Train Character Recognition Model

python reader_train.py

• Uses character-level dataset (LP reader yolov8/)
• Trains YOLOv8 to recognize individual characters
• Saves model to yolo_training/LPR_model*/weights/best.pt

⚙️ Configuration

Key Parameters in cropped_camera_detector.py:

• confidence_threshold: Minimum confidence for displaying detections (default: 0.5)
• crop_threshold: Minimum confidence for saving crops (default: 0.9)
• plate_model_path: Path to plate detection model
• char_model_path: Path to character recognition model

Model Paths:

• Plate Detection: yolo_training/vnlp_model7/weights/best.pt
• Character Recognition: yolo_training/LPR_model2/weights/best.pt

📊 Performance

Detection Model (vnlp_model7)

• Architecture: YOLOv8n
• Dataset: 1,005 Vietnamese license plate images
• Classes: 1 (license plate)

Character Recognition Model (LPR_model2)

• Architecture: YOLOv8n
• Classes: Alphanumeric characters (A-Z, 0-9)
• Purpose: Extract text from cropped license plates

🔍 Pipeline Workflow

1. Input: Live camera feed or image
2. Detection: Plate detection model identifies license plate regions
3. Cropping: High-confidence detections are cropped and saved
4. Recognition: Character recognition model reads text from cropped plates
5. Output: Annotated frames with license plate text

📋 Requirements

ultralytics>=8.0.0
opencv-python>=4.5.0
torch>=1.9.0
torchvision>=0.10.0
numpy>=1.21.0
pillow>=8.0.0

🐛 Troubleshooting

Common Issues:

1. Model not found: Ensure trained models exist in yolo_training/ directory
2. CUDA errors: Install CUDA-compatible PyTorch version
3. Camera not detected: Check camera permissions and connections
4. Low accuracy: Retrain models with more diverse datasets

Performance Optimization:

• GPU: Use gpu_detector.py for faster batch processing
• CPU: Reduce batch size for memory constraints
• Confidence: Adjust thresholds based on accuracy requirements

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Commit your changes
4. Push to the branch
5. Create a Pull Request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• Ultralytics YOLO for the base detection framework
• Vietnamese license plate dataset contributors
• OpenCV community for computer vision tools

📞 Contact

For questions or support, please open an issue in this repository.

---

Note: This system is designed for educational and research purposes. Ensure compliance with local privacy laws when using with live camera feeds.