Multiple Distribution Shift - Aerial (MDS-A): A Dataset for Test-Time Error Detection and Model Adaptation

Noel Ngu, Aditya Taparia, Gerardo I. Simari, Mario Leiva, Jack Corcoran, Ransalu Senanayake, Paulo Shakarian, Nathaniel D. Bastian · 2025

Read Multiple Distribution Shift - Aerial (MDS-A): A Dataset for Test-Time Error Detection and Model Adaptation by Noel Ngu, Aditya Taparia, Gerardo I. Simari, Mario Leiva, Jack Corcoran, Ransalu Senanayake, Paulo Shakarian, Nathaniel D. Bastian (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-2502-13289.txt.

Description

Models trained on clear skies stumble in fog, and a 2025 arXiv paper by Noel Ngu and co-authors releases the data to prove and fix it. Their collection holds simulated aerial image sets of the same scenes under six weather conditions, plus mixed weather test sets that differ sharply from any single training set.

Six baseline detectors ship with the data, so researchers can watch each one's accuracy slide as the weather drifts. The team also tries recent error detection techniques meant to flag and adapt to the shift. The dataset is public, aimed at test time reliability research.

Catalog record

CREATORNoel Ngu, Aditya Taparia, Gerardo I. Simari, Mario Leiva, Jack Corcoran, Ransalu Senanayake, Paulo Shakarian, Nathaniel D. Bastian
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE15 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-13289

Browse the full library · About the Tower of Babel