Dataset · ibm-esa-geospatial
TerraMesh Geospatial Satellite Dataset
ibm-esa-geospatial/TerraMesh
Over 9 million co-registered satellite imagery and geospatial data patches by IBM and ESA.
What's inside
A large-scale planetary-level multimodal dataset for training Earth observation foundation models. It combines over 9 million spatially co-registered patches from Sentinel-1 radar data, Sentinel-2 optical imagery, Copernicus DEM elevation models, NDVI indices, and land cover maps.
What you can do with it
- Pre-train foundation models for remote sensing of the Earth
- Feature extraction and multimodal analysis of satellite data
- Land cover classification and environmental monitoring
- Tasks
- Image feature extraction
- Size
- 1M–10M rows
- Downloads this month
- 15.5K
- Likes
- 35
Models trained on it
- ibm-esa-geospatial/TerraMind-1.0-base 83.3K
- ibm-esa-geospatial/TerraMind-1.0-Tokenizer-LULC 30.9K
- ibm-esa-geospatial/TerraMind-1.0-Tokenizer-S2L2A 28.8K
- ibm-esa-geospatial/TerraMind-1.0-Tokenizer-DEM 13.2K
- ibm-esa-geospatial/TerraMind-1.0-large 10.1K
- ibm-esa-geospatial/TerraMind-1.0-Tokenizer-S1RTC 7K
- ibm-esa-geospatial/TerraMind-1.0-tiny 5.6K
- ibm-esa-geospatial/TerraMind-1.0-Tokenizer-S1GRD 3.2K
How to load it
from datasets import load_dataset
ds = load_dataset("ibm-esa-geospatial/TerraMesh")