Multi-Modal Geospatial Datacubes from Five Satellite Sensors

Core-Five is a multi-modal geospatial dataset built for foundation models, unifying Earth Observation data from five essential sensors into aligned spatiotemporal datacubes. It includes optical Sentinel-2 data at 10m resolution and other sensor data for multi-modal vision tasks.

Use Cases

• Train multi-modal vision models on perfectly aligned spatiotemporal datacubes from five sensors for geospatial foundation model development.
• Apply contrastive learning across harmonized Sentinel-2 optical and other sensor modalities for self-supervised representation learning.
• Perform image segmentation or object detection on unified 10m resolution optical data from Sentinel-2 within the datacube structure.
• Use the aligned multi-sensor datacubes for tasks like super-resolution or image-to-image translation between different sensor modalities.

Strengths

• Unifies data from five essential Earth Observation sensors into a single, harmonized resource.
• Provides perfectly aligned spatiotemporal datacubes, a key feature for multi-modal model training.
• Includes Sentinel-2 optical data at a specified 10-meter spatial resolution.

Limitations

• Specific dataset scale, temporal coverage, and geographic scope are not provided in the input.
• The absence of detailed column definitions and sample data limits understanding of the exact data structure.

Provenance

Source

huggingface

Collection Method

Built from Earth Observation data unified from five satellite sensors.

Time Range

null

Freshness

Last updated on 2025-05 21.

Geography

Region is tagged as 'us', but specific spatial coverage is unknown.