Reproduce “GeoBiaset: A Counterfactual Benchmark for Demographic Bias in World-Level Geolocalization” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Niccolò Niccoli, Silvia Dani, Marco Mistretta, Andrew D. Bagdanov, Marco Bertini, Lorenzo Seidenari
GeoBiaset tests whether image geolocalization models rely on inserted people’s apparent demographic attributes. It pairs 329 geotagged backgrounds with synthetic edits that change foreground people while keeping locations fixed. The benchmark measures accuracy and directional prediction shifts. On GeoCLIP, continent-level accuracy fell from 81.9% to 55.3%; paired masking results suggest occlusion and background drift do not explain most of the loss.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.