So2Sat LCZ42 v4: Local Climate Zone Classification from Sentinel-1 and Sentinel-2

Provider
Xiao Xiang Zhu et al., Technical University of Munich (TUM) / DLR
License
CC-BY-SA-4.0 (source)
Size
10 files
Format
HDF5 (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/zhu-xlab/So2Sat-LCZ42)
Access
Free and public. No account with the data provider is needed.
Category
Earth Observation & Geospatial
Source
https://doi.org/10.1109/MGRS.2020.2964708

About this dataset

Benchmark for classifying Local Climate Zones (LCZ), the 17-class urban and natural land scheme used in urban climate research, from co-registered Sentinel-1 SAR and Sentinel-2 multispectral image patches. Patches are 32 x 32 pixels with 8 Sentinel-1 channels and 10 Sentinel-2 bands, labelled by hand. This copy of version 4 pairs the culture-10 split with geolocation: training covers 42 cities worldwide, validation the western half and testing the eastern half of 10 further cities spanning 10 cultural zones. The listing (measured 2026-11-14) shows a v4/ folder with training.h5.gz (about 16 GB), validation.h5.gz and testing.h5.gz (about 1.1 GB each), matching *_geo.h5 files with UTM coordinates, EPSG codes and world-file parameters, and read_file.py; about 18.2 GB in all. Gunzip, then read the sen1, sen2 and one-hot label arrays with h5py and numpy; the zhu-xlab GitHub repository has a notebook to write GeoTIFFs. TUM mediaTUM publishes version 4 under CC-BY-SA-4.0; cite Zhu et al., IEEE GRSM 2020.

How to load this dataset in Python

from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id="zhu-xlab/So2Sat-LCZ42", filename="v4/testing_geo.h5", repo_type="dataset")
print(path)

Data structure

File types in sampled files: HDF5

Files

Example file: v4/testing_geo.h5

Analyze this dataset without downloading it

On Scigantic, So2Sat LCZ42 v4: Local Climate Zone Classification from Sentinel-1 and Sentinel-2 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