Multi-Sensor Satellite Image and Text Dataset for Earth Observation

464,044 co-registered image-text pairs from Sentinel-1 and Sentinel-2 satellites form this large-scale dataset. It was created by BIFOLD-BigEarthNetv2-0 to advance vision-language learning for remote sensing data. The dataset was last updated on the platform in April 2026.

Use Cases

• Train image captioning models using the paired Sentinel-2 optical images and their textual descriptions.
• Develop multi-modal classifiers that fuse Sentinel-1 SAR data and Sentinel-2 multispectral data for land cover analysis.
• Benchmark visual question answering systems on geospatial imagery using the associated text prompts.
• Create contrastive learning models that align representations between satellite image patches and their text labels.

Strengths

• Contains 464,044 data points, providing a substantial scale for model training.
• Integrates multi-sensor data from both Sentinel-1 (SAR) and Sentinel-2 (optical) satellites.

Limitations

• Specific row counts, column details, and sample sizes for individual modalities are not provided.
• The geographic coverage is listed only as 'us', which may indicate a regional bias.

Provenance

Source

BIFOLD-BigEarthNetv2-0 on Hugging Face.

Collection Method

Co-registered image-text pairs collected from Sentinel-1 and Sentinel-2 satellite sensors.

Time Range

null

Freshness

null

Geography

Region listed as 'us' (United States).