Sentinel-2 Cloud Cover Segmentation Dataset (DrivenData challenge, expert-validated)

Provider
Radiant Earth Foundation
License
CC-BY-4.0 (source)
Format
JSON (in sampled files)
Stored at
Azure Blob Storage, https://radiantearth.blob.core.windows.net/mlhub/ref_cloud_cover_detection_challenge_v1
Access
Free and public. No account with the data provider is needed.
DOI
10.34911/rdnt.hfq6m7
Category
Earth Observation & Geospatial
Source
https://doi.org/10.34911/rdnt.hfq6m7

About this dataset

Pixel-level cloud masks for Sentinel-2 L2A imagery, first produced through the DrivenData 'On Cloud N' crowdsourcing competition and later reviewed and corrected by a team of expert annotators at Radiant Earth Foundation. The training data hold 22,728 chips of 512 x 512 pixels at 10 m sampled from scenes acquired between 2018 and 2020 across Africa, South America and Australia; each chip has the B02, B03, B04 and B08 bands (blue, green, red, near infrared) as separate GeoTIFFs and a binary cloud / no-cloud label raster. Measured 2026-11-15: more than 40,000 files (at least 21 GB listed) under final/public/ (train_features/ /B0x.tif, train_labels/, train_labels.zip, train_metadata.csv with location and acquisition time) and final/private/ (the held-out test chips), plus chip_id_crosswalk.csv mapping competition ids to Sentinel-2 scenes. Read with rasterio; DrivenData's PyTorch Lightning benchmark uses the Planetary Computer for extra bands. Useful for cloud masking, compared with Fmask and s2cloudless. Released under CC BY 4.0; cite Radiant Earth Foundation (2022), Sentinel-2 Cloud Cover Segmentation Dataset, doi:10.34911/rdnt.hfq6m7.

How to access this dataset in Python

import adlfs
fs = adlfs.AzureBlobFileSystem(account_name='radiantearth', anon=True)
fs.ls('mlhub/ref_cloud_cover_detection_challenge_v1/')[:10]

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

Data structure

File types in sampled files: JSON

Files

Example file: .source/metadata.json

Analyze this dataset without downloading it

On Scigantic, Sentinel-2 Cloud Cover Segmentation Dataset (DrivenData challenge, expert-validated) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Earth Observation & Geospatial datasets