Copernicus-Bench: 15 Downstream Tasks Across Sentinel-1, -2, -3 and -5P for EO Foundation Models
- Provider
- Wang et al. (Technical University of Munich, DLR)
- License
- CC-BY-4.0 (source)
- Size
- 49 files
- Format
- HDF5 (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/wangyi111/Copernicus-Bench)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Earth Observation & Geospatial
- Source
- https://arxiv.org/abs/2503.11849
About this dataset
Copernicus-Bench is the evaluation suite released with the Copernicus-FM foundation model (Wang et al., 2025) to test Earth-observation foundation models on every major Copernicus Sentinel mission. It holds 15 hierarchical tasks in three levels. Level 1, preprocessing: cloud segmentation on Sentinel-2 (from CloudSEN12) and Sentinel-3 OLCI (new). Level 2, base applications: EuroSAT land-use classification on S1 and S2, BigEarthNet v2 multi-label classification on S1 and S2, DFC2020 land cover segmentation on S1 and S2, and Sentinel-3 land cover classification and segmentation against Copernicus Global Land Cover LC100 (new). Level 3, specialised applications: Kuro Siwo flood change detection on S1, So2Sat LCZ42 local climate zone classification on S2, Sentinel-3 aboveground biomass regression, and Sentinel-5P NO2 and O3 air-quality regression (new). Each task sits in its own folder (l1_cloud_s2, l2_bigearthnet_s1s2, l3_airquality_s5p and so on), about 155 GB in all. The card lists each task's licence: CC BY 4.0 for the new and most derived tasks, CC0 for cloud S2, MIT for EuroSAT-S2 and Kuro Siwo flood, CDLA-Permissive-1.0 for BigEarthNet. Cite Wang et al. 2025.
How to load this dataset in Python
from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id="wangyi111/Copernicus-Bench", filename="deprecated/so2sat_s1s2/culture_10/test-new.h5", repo_type="dataset")
print(path)Data structure
File types in sampled files: HDF5
Files
Example file: deprecated/so2sat_s1s2/culture_10/test-new.h5
Analyze this dataset without downloading it
On Scigantic, Copernicus-Bench: 15 Downstream Tasks Across Sentinel-1, -2, -3 and -5P for EO Foundation Models 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
- Merged Google-Microsoft Buildings
- Data to support downscaling regional economic activity
- NASA HEASARC: SAS-2 Mission Data Archive
- Satellite - Ocean Colour - MODIS - 1 day - Net Primary Productivity (OC3 model and Eppley-VGPM algorithm)
- SPHEREx Quick Release (QR): An All-Sky Spectral Survey
- CanElevation - LiDAR Point Clouds
- World Settlement Footprint
- HLS Sentinel-2 Multi-spectral Instrument Surface Reflectance Daily Global 30m v2.0