---
pretty_name: PhilEO-Bench Building Density Regression
task_categories:
  - other
annotations_creators:
  - other
source_datasets:
  - generated
tags:
  - image
  - geospatial
  - remote-sensing
  - earth-observation
  - satellite-imagery
  - phi-sat-2
  - simulation
  - regression
  - building-density
---

# Simulated PhiSat Bench Dataset - Buildings

This repository contains a simulated dataset derived from Sentinel-2 data for building analysis. 
Specifically, the dataset simulates outputs from the PhiSat2 satellite. 

## Label Description

Each sample in the dataset includes a single-channel label. 
The labels are stored as floating-point values that represent the estimated percentage of building coverage within each pixel. 
For a pixel with a 10-meter resolution (representing 100 square meters), the label values range from **0 to 100**, where:
- **0** indicates no building coverage, and
- **100** indicates full building coverage.

## Use Case

This dataset is intended for training models to predict building coverage as continuous values based on satellite imagery.

<!-- phi2fm-dataset-figures:start -->
## Dataset Visual Summary

The figures below are designed for quick visual inspection rather than quantitative evaluation. Each representative panel shows the source image, a label overlay, and an adaptive heatmap that makes sparse targets easier to inspect.

- Samples: `676`
- Image shape: `(8, 5263, 3797)`
- Label shape: `(1, 5263, 3797)`
- Approximate mean label coverage: `0.024%`
- Approximate mean positive-pixel fraction: `0.0573`

![Representative samples](./figures/building_preview_grid.png)

![Label distribution summary](./figures/building_label_summary.png)
<!-- phi2fm-dataset-figures:end -->