A large-scale benchmark dataset derived from Sentinel-1 radar and Sentinel-2 optical satellite imagery. The dataset, created by GFM-Bench, is pre-processed with images upsampled to 120x120 pixels and land cover labels mapped to 19 categories. It was last updated on the platform in October 2025.
Use Cases
- Multi-label land cover classification based on Sentinel-1 and Sentinel-2 satellite imagery
- Benchmarking computer vision models on a large-scale remote sensing dataset
- Developing multimodal fusion models using combined radar and optical data
- Analyzing land use patterns from pre-processed 120x120 pixel image patches
Strengths
- Derived from two distinct satellite sources: Sentinel-1 (radar) and Sentinel-2 (optical) imagery
- Pre-processed with images upsampled to a consistent 120x120 pixel resolution
- Land cover labels are mapped to a consolidated set of 19 categories from an original 43
Limitations
- Column-level documentation is absent; field semantics must be inferred after download
- Row count is unknown, which may limit suitability assessment
- Description metadata is limited; actual data quality requires manual inspection after download
Provenance
- Source
- GFM-Bench
- Collection Method
- Derived and pre-processed from Sentinel-1 and Sentinel 2 satellite imagery
- Time Range
- null
- Freshness
- Last updated 2025-10-29 21:31:41; freshness should be verified
- Geography
- null