Coralguard/Corals2

🪸 Coral Health Analysis System

An AI-powered system for coral species classification and health assessment using deep learning and color analysis.

🎯 Features

1. Coral Species Classification

• —Uses EfficientNet-B0 CNN model
• —Classifies into 4 coral types:
• —🪨 Boulder Coral
• —🌿 Branched Coral
• —🍽️ Plate Coral
• —🌊 Soft Coral
• —Displays confidence scores for each class

2. Health Assessment

• —CoralWatch color-based health scoring
• —Analyzes both dark and light color patches
• —Provides health status:
• —✅ Healthy: Average score > 3
• —⚠️ Stressed/Bleached: Average score ≤ 3

3. Advanced Segmentation

• —Removes water, sand, and background
• —Isolates coral regions for accurate analysis
• —Multi-strategy preprocessing

4. Visual Overlays

• —Color-coded health visualization
• —Segmentation masks
• —Confidence charts

🚀 How to Use

1. 1.Upload a coral image (JPG, PNG)
2. 2.Click "Analyze Coral"
3. 3.View results:
4. 4.Species classification
5. 5.Health status
6. 6.Color analysis
7. 7.Visual overlays

🔬 Technology Stack

• —Deep Learning: PyTorch, EfficientNet-B0
• —Computer Vision: OpenCV, scikit-learn
• —Interface: Gradio
• —Color Analysis: CoralWatch methodology

📊 Model Details

• —Architecture: EfficientNet-B0
• —Input Size: 224x224
• —Training: ImageNet pretrained + fine-tuned on coral dataset
• —Classes: 4 coral species

🎨 Health Scoring

Based on CoralWatch color reference chart:

• —Score 1-2: Severely bleached (lightest)
• —Score 3: Moderately stressed
• —Score 4-6: Healthy (darkest)

📝 Citation

If you use this system in your research, please cite:

Coral Health Analysis System
Powered by EfficientNet-B0 and CoralWatch methodology

🤝 Contributing

Contributions are welcome! Feel free to:

• —Report issues
• —Suggest improvements
• —Add new features

📄 License

MIT License

Built with ❤️ for coral reef conservation