Probabilistic Machine Learning for Noisy Labels in Earth Observation

Spyros Kondylatos, Nikolaos Ioannis Bountos, Ioannis Prapas, Angelos Zavras, Gustau Camps-Valls, Ioannis Papoutsis · 2025

Read Probabilistic Machine Learning for Noisy Labels in Earth Observation by Spyros Kondylatos, Nikolaos Ioannis Bountos, Ioannis Prapas, Angelos Zavras, Gustau Camps-Valls, Ioannis Papoutsis (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-03478.txt.

Description

Satellite labels lie often enough to mislead the models trained on them, and Spyros Kondylatos and five co-authors make the models admit their doubts in a 2025 paper. Labels in Earth observation come from photo interpreters, old maps, and crowds, each with its own errors, and ordinary models swallow the errors whole.

The team's probabilistic models instead learn how noise depends on the input and report uncertainty beside each prediction. They trained such models across a spread of high-stakes tasks, flood and crop mapping among them, with varied sensors and setups, and built a pipeline to grade both accuracy and honesty. The uncertainty-aware models beat standard ones on most datasets and measures, and their stated doubts checked out under scrutiny. The authors argue trust, not just accuracy, is the deliverable for Earth observation.

Catalog record

CREATORSpyros Kondylatos, Nikolaos Ioannis Bountos, Ioannis Prapas, Angelos Zavras, Gustau Camps-Valls, Ioannis Papoutsis
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE58 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-03478

Browse the full library · About the Tower of Babel