lloyd97/mauritius-landcover

--- title: Mauritius Land Cover Classification emoji: 🌍 colorFrom: green colorTo: blue sdk: docker app_port: 7860 pinned: false --- # Mauritius Land Cover Classification Deep learning-based land cover classification for Mauritius using Sentinel-2 satellite imagery and U-Net with ResNet50 encoder. ## Overview This project uses a U-Net segmentation model with a ResNet50 encoder to classify land cover types across Mauritius from Sentinel-2 multispectral satellite imagery. ### Land Cover Classes (Apple Maps Style) | Class | Color | RGB | Description | |-------|-------|-----|-------------| | Water | Soft Blue | (168, 216, 234) | Ocean, lagoons, rivers, reservoirs | | Forest | Rich Green | (139, 195, 74) | Native forests, dense vegetation | | Plantation | Light Green | (197, 225, 165) | Sugarcane fields, agricultural land | | Urban | Warm Gray | (215, 204, 200) | Buildings, developed areas | | Roads | Medium Gray | (158, 158, 158) | Highways, streets, paved surfaces | | Bare Land | Sandy Cream | (239, 235, 233) | Quarries, beaches, cleared land | ### Model Performance | Metric | Value | |--------|-------| | **Overall Accuracy** | 73.50% | | **Water Accuracy** | 99.77% | | **Forest Accuracy** | 88.80% | | **Urban Accuracy** | 85.29% | | **Bare Land Accuracy** | 76.82% | | **Plantation Accuracy** | 54.75% | | **Roads Accuracy** | 37.65% | ## Features - **Live Interactive Map**: Real-time land cover classification as you pan across Mauritius - **9-Band Input**: Uses B2, B3, B4, B8, B11, B12 + NDVI, NDWI, NDBI indices - **Apple Maps-inspired styling**: Clean, aesthetic color palette - **Google Earth Engine integration**: Automatic Sentinel-2 imagery download ## Installation ```bash # Clone the repository git clone github.com cd mauritius-landcover # Create virtual environment python -m venv venv source venv/bin/activate # Linux/Mac # or: venv\Scripts\activate # Windows # Install dependencies pip install -r requireme …