GeoSpatial ML

We trained a logistic regression model that reaches 96.04% test accuracy on EuroSAT with 306 learned parameters by using 33 fixed spectral and spatial measurements of each patch.

Between the two of us, we reviewed 22 paper submissions this summer, spread across NeurIPS, WACV, and TerraBytes (a geospatial workshop at ECCV).

What colors does the Earth actually come in?

The broad availability of multispectral data is one thing that separates satellite-based machine learning from mainstream computer vision.

Disclaimer: nobody should deploy a neural network this way.

Multi-temporal super-resolution of a Sentinel-2 scene.

The abundance of open satellite imagery and advances in geospatial ML and remote sensing methods have made it possible to monitor a growing list of variables directly from orbit.

In our TerraBit post last time, we binary-quantized the global Clay v1.5 embeddings down to 128 bytes per patch and served 50M of them from static object storage — implementing planetary scale retrieval entirely in the browser.

Unfinished business Last time, we compressed earth embeddings 64× with less than 2% loss on patch classification.

Update (2026-03-26): OlmoEarth-nano results throughout have been recomputed with properly normalized inputs.

How well do segmentation models actually use long-range spatial information to make decisions?

How can you join AEF embeddings to census blocks, and how well do they predict different variables?

One notebook, a few hundred lines of Python, and you go from raw Sentinel-2 imagery to a georeferenced water map you can open in QGIS.

Welcome to GeoSpatial ML — a place to share what we’re exploring, building, and reading at the intersection of geospatial data and machine learning.