S2Wc Rss Like: Weakly Supervised Land Cover Segmentation

ESA WorldCover 2020 v100 dataset provides a global 10-meter resolution land cover map. This dataset is derived from it for the task of weakly supervised semantic segmentation. The dataset was created by author j-h-f and last updated on the Hugging Face platform in April 2025.

Use Cases

• Training weakly supervised semantic segmentation models based on the derived land cover labels.
• Developing methods for land use classification based on satellite imagery.
• Benchmarking computer vision algorithms for geospatial analysis tasks.

Strengths

• Derived from the authoritative ESA WorldCover 2020 v100 dataset, which has a global scope.
• Last updated on 2025-04-30, indicating recent maintenance on the platform.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and sample data are unknown, which may limit suitability assessment.
• License information is unknown, which may restrict commercial use.

Provenance

Source

Derived from the ESA WorldCover 2020 v100 dataset (Zanaga et al., 2021).

Collection Method

Method of derivation for weakly supervised learning is not specified in the provided metadata.

Time Range

Based on the 2020 reference data from the source dataset.

Freshness

Last updated 2025-04-30 10:07:22; freshness should be verified.

Geography

Global, as the source ESA WorldCover dataset provides worldwide coverage.