GTSRB Traffic Sign Detection - CORE Module

This is a streamlined core module for German Traffic Sign Recognition using TensorFlow Lite. It provides simple Python scripts for detecting traffic signs and outputting results to JSON format.

📁 Directory Structure

CORE/ ├── traffic_sign_detector.py # Main detection class ├── single_detect.py # Single image detection script ├── batch_detect.py # Batch processing script ├── requirements.txt # Python dependencies ├── models/ # Model files │ └── gtsrb_model.lite # TensorFlow Lite model ├── test_images/ # Sample test images │ ├── 00000.png │ ├── 00001.png │ └── ... └── output/ # Detection results └── detection_results.json

🚀 Quick Start

1. Install Dependencies

bash pip install -r requirements.txt

2. Single Image Detection

bash python single_detect.py test_images/00000.png

With JSON output: bash python single_detect.py test_images/00000.png -o output/single_result.json

3. Batch Processing

bash python batch_detect.py test_images/

With custom output file: bash python batch_detect.py test_images/ -o output/my_results.json

4. Using the Detection Class Directly

```python from trafficsigndetector import TrafficSignDetector

Initialize detector

detector = TrafficSignDetector()

Detect single image

result = detector.detectsign('testimages/00000.png') print(result)

Batch detection

results = detector.detect_batch(['image1.jpg', 'image2.jpg'])

Save to JSON

detector.saveresultsto_json(results, 'output/results.json') ```

🎯 Features

• Simple API: Easy-to-use Python classes and functions
• JSON Output: Structured results with confidence scores and metadata
• Batch Processing: Process multiple images efficiently
• Configurable: Adjustable confidence thresholds
• Detailed Results: Top predictions, inference times, and model info
• Error Handling: Robust error handling and logging

📊 Output Format

The detection results are saved in JSON format with the following structure:

json { "detection_summary": { "total_images": 5, "successful_detections": 4, "failed_detections": 1, "success_rate": 80.0, "average_inference_time_ms": 45.32, "detection_timestamp": "2025-06-26T10:30:00" }, "detections": [ { "image_path": "test_images/00000.png", "timestamp": "2025-06-26T10:30:00", "inference_time_ms": 42.5, "detected": true, "primary_detection": { "class_id": 14, "label": "Stop", "confidence": 0.9876 }, "top_predictions": [ { "class_id": 14, "label": "Stop", "confidence": 0.9876 }, { "class_id": 13, "label": "Yield", "confidence": 0.0123 } ], "model_info": { "model_path": "models/gtsrb_model.lite", "confidence_threshold": 0.3, "input_shape": [1, 224, 224, 3], "total_classes": 43 } } ] }

🏷️ Traffic Sign Classes

The model can detect 43 different German traffic sign classes:

• Speed Limits: 20, 30, 50, 60, 70, 80, 100, 120 km/h
• Warning Signs: Dangerous curves, road work, pedestrians, etc.
• Mandatory Signs: Turn directions, keep right/left, etc.
• Prohibition Signs: No overtaking, no entry, etc.
• Priority Signs: Right-of-way, yield, stop, etc.

⚙️ Configuration Options

Command Line Arguments

Single Image Detection: - image_path: Path to the image file (required) - -o, --output: Output JSON file path - -m, --model: Path to TensorFlow Lite model - -c, --confidence: Confidence threshold (default: 0.3)

Batch Processing: - input_dir: Directory containing images (required) - -o, --output: Output JSON file path (default: output/batch_results.json) - -m, --model: Path to TensorFlow Lite model - -c, --confidence: Confidence threshold (default: 0.3)

TrafficSignDetector Class

python detector = TrafficSignDetector( model_path='models/gtsrb_model.lite', # Path to model confidence_threshold=0.3 # Confidence threshold )

🔧 Requirements

• Python 3.7+
• TensorFlow 2.10+
• NumPy 1.21+
• Pillow 8.3+

📝 Usage Examples

Example 1: Quick Detection

bash cd CORE python single_detect.py test_images/00000.png

Example 2: Batch with Custom Threshold

bash python batch_detect.py test_images/ -c 0.5 -o output/high_confidence.json

Example 3: Custom Model

bash python single_detect.py image.jpg -m /path/to/custom_model.lite

🐛 Troubleshooting

Model not found error: - Ensure gtsrb_model.lite is in the models/ directory - Check file permissions

Import errors: - Install required packages: pip install -r requirements.txt - Ensure TensorFlow is properly installed

No detections: - Try lowering the confidence threshold with -c 0.1 - Check if the image contains traffic signs - Verify image format (PNG, JPG, etc.)

📈 Performance

• Inference Speed: ~20-50ms per image (CPU)
• Memory Usage: ~100MB (model loaded)
• Accuracy: High accuracy on German traffic signs
• Supported Formats: PNG, JPG, JPEG, BMP, TIFF