DATA_SUMMARY.md

# Satellite Image Dataset - Data Summary

<!--
Satellite Image Dataset Project
Created by: RSK World
Website: https://rskworld.in
Email: help@rskworld.in
Phone: +91 93305 39277
-->

## Generated Sample Data

This document summarizes the sample data files created for the Satellite Image Dataset project.

## Data Files Created

### Land Cover Labels (`data/labels/`)
Contains JSON files with land cover classification annotations:
- `sample_001.json` through `sample_010.json` (10 files)
- Each file includes:
- Image ID
- Land cover classes (water, forest, urban, agriculture, barren, grassland)
- Polygon coordinates for each region
- Area calculations
- Confidence scores
- Annotation metadata

### Building Detection (`data/building_detection/`)
Contains JSON files with building detection annotations:
- `sample_001.json` through `sample_010.json` (10 files)
- Each file includes:
- Image ID
- Building bounding boxes
- Confidence scores
- Building types (residential, commercial, industrial)
- Height estimates
- Area calculations

### Geospatial Metadata (`data/metadata/`)
Contains JSON files with geospatial metadata:
- `sample_001.json` through `sample_010.json` (10 files)
- Each file includes:
- Image ID
- Coordinate Reference System (CRS)
- Geographic bounds (latitude/longitude)
- Resolution
- Acquisition date
- Sensor information (Landsat 8/9, Sentinel-2, MODIS)
- Spectral bands
- Cloud cover percentage
- Image dimensions
- Provider information

## Statistics

- **Total Sample Images**: 10
- **Total Data Files**: 30 (10 labels + 10 buildings + 10 metadata)
- **Land Cover Classes**: 6 (water, forest, urban, agriculture, barren, grassland)
- **Building Types**: 3 (residential, commercial, industrial)
- **Sensors**: Landsat 8, Landsat 9, Sentinel-2, MODIS

## Data Format

All data files are in JSON format with the following structure:

### Label Format
```json
{
"image_id": "sample_001",
"classes": ["water", "forest", "urban"],
"regions": [
{
"class": "urban",
"polygon": [[x1, y1], [x2, y2], ...],
"area": 10000.0,
"confidence": 0.95
}
]
}
```

### Building Detection Format
```json
{
"image_id": "sample_001",
"buildings": [
{
"id": "bld_001",
"bbox": [x, y, width, height],
"confidence": 0.95,
"area": 2500.0,
"type": "residential",
"height_estimate": 15.5
}
]
}
```

### Metadata Format
```json
{
"image_id": "sample_001",
"crs": "EPSG:4326",
"bounds": {
"min_lon": -122.5,
"min_lat": 37.5,
"max_lon": -122.0,
"max_lat": 38.0
},
"resolution": 0.5,
"acquisition_date": "2024-01-15",
"sensor": "Landsat 8"
}
```

## Usage

To use this data:

1. **Load labels:**
```python
from data_loader import SatelliteDatasetLoader
loader = SatelliteDatasetLoader()
image, labels = loader.load_image_pair("sample_001")
```

2. **Load building detections:**
```python
buildings = loader.load_building_detections("sample_001")
```

3. **Load metadata:**
```python
metadata = loader.load_metadata("sample_001")
```

## Generating More Data

To generate additional sample data files, run:

```bash
python generate_sample_data.py
```

You can modify the `num_samples` parameter in the script to generate more or fewer samples.

## Notes

- All sample data is synthetic and generated for demonstration purposes
- Geographic coordinates are based on the San Francisco Bay Area region
- Building and land cover annotations are randomly generated but follow realistic patterns
- All files include RSK World attribution

---

**Created by: RSK World**
Website: https://rskworld.in
Email: help@rskworld.in
Phone: +91 93305 39277