AI BriefingAi2 Unveils OlmoEarth Platform
Key point
Ai2 has unveiled a geospatial foundation model and inference platform that processes large-scale satellite data.
Details
The Allen Institute for AI (Ai2) announced OlmoEarth, a family of geospatial observation foundation models pretrained on about 10TB of multimodal satellite data, along with a dedicated platform for it.
The OlmoEarth platform goes beyond simply providing models—it's designed to let environmental organizations run large-scale inference without building their own infrastructure. Key features include:
- Large-scale inference performance: Completes inference over continent-scale areas in about a day, processing tens of terabytes of imagery at a very low cost per square kilometer.
- Engineering optimization: Features a pipeline that automatically aligns and processes satellite imagery across various projections, resolutions, and sensor types.
- Efficient resource allocation: Maximizes cost efficiency through a hardware optimization strategy that allocates CPUs for data download and preprocessing and GPUs for model execution.
The platform is intended to be used by mission-driven organizations to derive practical insights in areas such as deforestation monitoring, food security, and wildfire risk prediction.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
NASA-IBM Releases Open-Source Multimodal Geospatial AI Model 'Lunar Foundation' for Lunar Exploration