Audit crop-map accuracy under biased field sampling: deployment metrics with confidence intervals from convenience labels, via prediction-powered inference.
-
Updated
Nov 3, 2026 - Python
Audit crop-map accuracy under biased field sampling: deployment metrics with confidence intervals from convenience labels, via prediction-powered inference.
An end-to-end GeoAI platform for agricultural monitoring using satellite foundation model embeddings, AlphaEarth, Galileo representations, NDVI-based vegetation analysis, and machine learning based crop classification.
A web application created to predict the crop yield based on historical data. It can perform basic analysis, along with plotting the crop harvest in various states.
This study presents the first comprehensive review of large-scale, pixel-wise crop mapping methods. It integrates novel systematic experiments that evaluate the classification performance of entire workflows and assess their spatial, temporal, and sensor transferability across five diverse agricultural sites in the U.S. and Serbia.
Code and aggregate analysis artifacts for reliability auditing of confidence-based early-season crop mapping across regions and years.
Part of NASA Harvest's RISE. Sat imagery download, training data creation, model deploy, STAC + COGs output. Can be run locally and on Azure ML.
Preregistered test of AlphaEarth satellite embeddings against Sentinel-2 for crop mapping, with labels the model saw (California) and independent labels (France), under random and spatial cross-validation.
To associate your repository with the crop-mapping topic, visit your repo's landing page and select "manage topics."