← Directory
software · Global
Prithvi-EO-2.0-300M
Hugging Face model ibm-nasa-geospatial/Prithvi-EO-2.0-300M
Claim this listing
Represent Prithvi-EO-2.0-300M? Send a claim for review so we can discuss listing updates. Claims do not automatically grant ownership or verification.
What's in it
- Data offered docs-verified
- Pretrained 300M-parameter Earth observation foundation model (ViT, masked autoencoder, 3D spatiotemporal embeddings) · Image reconstruction on multi-temporal HLS GeoTIFF stacks via inference.py · Backbone for fine-tuning in TerraTorch (prithvi_eo_v2_tiny_tl example)
- Resolution docs-verified
- 30 m (HLS V2 pretraining granularity)
- Formats docs-verified
- GeoTIFF · PyTorch model weights
- Access docs-verified
- Open licence: Apache-2.0; downloadable from Hugging Face. DOCS state no API key requirement; not deployed by any Inference Provider.
Useful for
- Use cases llm-inferred
- Landslide segmentation screening by fine-tuning on labelled imagery (TerraTorch landslide notebook pattern) · Multitemporal land-cover or crop-type mapping from HLS time series · Reconstruction or gap-filling of cloud-affected HLS image stacks before site-scale analysis
- Audience llm-inferred
- GIS analysts · remote sensing specialists · ecologists · geotechnical engineers
Limitations
- Constraints llm-inferred
- No hosted inference: not deployed by any Inference Provider, so it must be run locally or in your own pipeline · Inference expects six-band HLS-style reflectance GeoTIFFs in chronological order · Fine-tuning needs PyTorch/TerraTorch and GPU compute (example notebooks use a T4 GPU runtime) · Pretraining resolution is 30 m; suitability for finer imagery is not stated in DOCS
- Licence
- apache-2.0
- Authentication
- No API key required (repository README, auto-extracted; check upstream requirements)
- GitHub stars
- Unknown
- Last commit
- Unknown
- Last successful page check
- 14/10/2026
- Page health
- healthy
- Package
- Not recorded
- Verification
- Not independently verified
Provenance & evidence
Verified 2026-10-15 via HF API: exists; 19795 downloads/30d, 64 likes.
Metadata observed 14/10/2026 · research sweep: manual-2026-10-15 (Hugging Face models)