Multi-Modal Foundation · allenai
OlmoEarth v1.1 Base
A remote sensing foundation model for analysing Sentinel-1, Sentinel-2, and Landsat satellite imagery.
What OlmoEarth v1.1 Base is
OlmoEarth v1.1 Base is an Earth-observation foundation model developed by the Allen Institute for AI (Ai2). It is designed to interpret remote sensing imagery for applications across Earth science, climate study, and environmental conservation.
The checkpoint represents an updated base-scale release that provides improved pretraining and downstream performance compared to its initial version. It functions as a shared visual backbone that researchers can adapt to various geospatial analysis workflows without training vision networks from scratch.
How it works
The model employs a Vision Transformer Base (ViT-Base) architecture containing 114 million parameters. It processes single satellite scenes or multi-temporal image series derived from multiple optical and radar earth observation platforms, specifically Sentinel-1, Sentinel-2, and Landsat.
During deployment, the network extracts spatial and temporal feature representations from satellite inputs. These representations can be extracted directly as fixed numerical embeddings or adapted through downstream supervised fine-tuning.
Key facts
- Architecture
- Vision Transformer Base (ViT-Base)
- Parameter count
- 114 million
- Supported sensors
- Sentinel-1, Sentinel-2, and Landsat
- Data format
- Single images and image time series
What you can use it for
- Satellite image segmentation
- Fine-tuning the backbone to delineate and segment land cover categories, surface water, or vegetation zones across geographical regions.
- Earth observation feature extraction
- Generating high-dimensional image embeddings from single scenes or temporal stacks for clustering and spatial analysis.
- Time series remote sensing
- Analysing multi-temporal satellite observations across seasons to observe environmental changes over time.
Run it here
We list this model but cannot run it yet: no pipeline_tag — needs a human.
Everything on this page still applies. Running it needs a loader for this repository, not a bigger machine.
Strengths and limits
Good at
- Processes data from both optical (Sentinel-2, Landsat) and radar (Sentinel-1) satellite constellations.
- Handles both individual static images and multi-temporal image time series.
- Provides improved pretraining representations and downstream task execution over the original v1 checkpoint.
Watch out for
- Downstream tasks such as semantic segmentation require task-specific fine-tuning before producing final outputs.
- Inputs are restricted to sensor types and modalities compatible with Sentinel-1, Sentinel-2, and Landsat imagery.
Inputs and outputs
You provide
Sequence or text
A protein or nucleotide sequence, or free text — whichever this model was trained on.
You get back
Mean-pooled representation, one vector per input.
Licence
Check the licence before commercial use · See model card
The See model card licence carries terms worth reading before you depend on it.
Questions
Can I use OlmoEarth v1.1 Base commercially?
Check the licence before commercial use. The See model card licence carries terms worth reading before you depend on it.
What satellite sensors can be used with OlmoEarth v1.1 Base?
The model is designed to work with remote sensing imagery from Sentinel-1, Sentinel-2, and Landsat platforms.
Can the model analyse temporal changes?
Yes, it accepts both single satellite scenes and multi-frame image time series.
What architecture does this model use?
It is built on a standard Vision Transformer Base (ViT-Base) architecture containing 114 million parameters.
How does v1.1 differ from the original OlmoEarth v1?
The v1.1 checkpoint features improved pretraining and enhanced performance across downstream evaluation tasks.
Written from OlmoEarth v1.1 Base's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.