# Semantic Segmentation of Crop Type in Ghana (Sentinel-1, Sentinel-2 and PlanetScope, 2016-2017)

## Facts

- Provider: Rustowicz, Cheong, Wang, Ermon, Burke and Lobell (Stanford University)
- License: CC-BY-SA-4.0 (https://radiantearth.blob.core.windows.net/mlhub/su-african-crops-ghana/README.md)
- Format: JSON (in sampled files)
- Stored at: Azure Blob Storage, https://radiantearth.blob.core.windows.net/mlhub/su-african-crops-ghana
- Access: public, readable without an account
- DOI: https://doi.org/10.34911/rdnt.ry138p
- Category: Earth Observation & Geospatial
- Source: https://doi.org/10.34911/rdnt.ry138p
- Page: https://scigantic.com/archive/cf028484-243c-5a0b-b0c2-a44b3be148a0
- Keywords: Earth Observation & Geospatial, image segmentation, satellite imagery, time series, JSON, Azure Blob Storage

## About

Pixel-level crop type labels for smallholder farms in Ghana with time series from three satellites, from Stanford University's Sustainability and AI Lab. Small irregular plots, frequent cloud and scarce ground truth make these regions hard for crop mapping, so each chip carries Sentinel-1 SAR backscatter, Sentinel-2 multispectral reflectance and high-frequency 3 m PlanetScope imagery across 2016 and 2017, with one crop label per pixel; pixels without ground truth are 0. The crop is assumed constant over each chip's time series. Measured 2026-11-15: more than 40,000 files (201 GB listed and more beyond). original/ holds NumPy arrays (s1_npy/, s2_npy/, planet_npy/, truth_npy/) as used in the paper; planet/, s1/, s2/ and truth/ hold the same data as GeoTIFFs with STAC JSON. Load the arrays with numpy or the GeoTIFFs with rasterio; the authors' 2D and 3D U-Net and CLSTM code is on GitHub (roserustowicz/crop-type-mapping). Released under CC BY-SA 4.0; cite Rustowicz et al. (2019), CVPR Workshops, and the Radiant MLHub DOI 10.34911/rdnt.ry138p.

## Load the data

```python
import adlfs
fs = adlfs.AzureBlobFileSystem(account_name='radiantearth', anon=True)
fs.ls('mlhub/su-african-crops-ghana/')[:10]
```

Lists the files with anonymous access. Runs on any machine, with no cloud account.

## Data structure

File types in sampled files: JSON

Example file: `.source/metadata.json`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/cf028484-243c-5a0b-b0c2-a44b3be148a0
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/cf028484-243c-5a0b-b0c2-a44b3be148a0
- Over MCP: `get_schema_card` with id `cf028484-243c-5a0b-b0c2-a44b3be148a0` at https://api.scigantic.com/mcp

## Related topics

- [Earth Observation & Geospatial datasets](https://scigantic.com/topics/earth-observation-and-geospatial)
- [JSON datasets](https://scigantic.com/topics/json)
- [Satellite imagery datasets](https://scigantic.com/topics/satellite-imagery)
- [Image segmentation datasets](https://scigantic.com/topics/image-segmentation)
- [Time series datasets](https://scigantic.com/topics/time-series)
