So2Sat-LCZ42: Satellite Imagery for Local Climate Zone Classification Across 52 Cities

A dataset pairing satellite imagery patches with local climate zone labels for urban areas. Training data covers 42 cities globally, with validation and testing data each covering halves of 10 other cities across 10 cultural zones. The dataset was created by zhu-xlab and last updated on Hugging Face in March 2026.

Use Cases

• Train image classification models to identify local climate zones based on paired satellite imagery patches.
• Validate remote sensing models using data from distinct cultural zones mentioned in the description.
• Benchmark geospatial machine learning algorithms on a dataset with a defined training/validation/testing split.
• Analyze urban land cover characteristics across different global cities.

Strengths

• Training data covers 42 cities from around the world.
• Validation and testing splits are derived from 10 other cities across 10 distinct cultural zones.
• Data includes paired sensor inputs (SEN1 and SEN2) as mentioned in the file description.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

zhu-xlab on Hugging Face.

Collection Method

Likely compiled from satellite imagery sources and labeled for local climate zones.

Time Range

null

Freshness

Last updated 2026-03-04 16:44:28; freshness should be verified.

Geography

52 cities around the world, including 10 cities covering 10 cultural zones.