# FairSAM: Fair Classification on Corrupted Image Data Through Sharpness-Aware Minimization Source: [FairSAM: Fair Classification on Corrupted Image Data Through Sharpness-Aware Minimization]() Collection index: [llms.txt]() Data updated: 2026-06-24T05:00:31.418Z Authors: Yucong Dai, Jie Ji, Xiaolong Ma, Yongkai Wu Published: 2025-03-29T01:51:59.000Z ## Abstract Image classification models trained on clean data often degrade sharply when exposed to corrupted test or deployment data, such as images with impulse noise, Gaussian noise, or environmental noise. This degradation reduces overall performance and disproportionately affects demographic subgroups, raising algorithmic bias concerns. Although robust learning algorithms such as Sharpness-Aware Minimization improve overall robustness and generalization, they do not address biased performance degradation across demographic subgroups. Existing fairness-aware machine learning methods reduce performance disparities but struggle to maintain robust and equitable accuracy across demographic subgroups under data corruption. This limitation reveals an inherent tension between robustness and fairness under corrupted data. To address these challenges, we introduce a metric to assess performance degradation across subgroups under data corruption. We propose FairSAM, a framework that integrates Fairness-oriented strategies into SAM to equalize performance across demographic groups under corrupted conditions. Experiments on multiple real-world datasets and prediction tasks show that FairSAM balances robustness and fairness in corrupted image classification. The framework yields a structured solution for fair and robust image classification in the presence of data corruption. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Gender Classification on FairFace level-5 Gaussian noise corrupted (test)](): Gender Classification; FairFace level-5 Gaussian noise corrupted (test); Accuracy (s+): 63.7. - [Pleural Effusion Classification on CheXpert](): Pleural Effusion Classification; CheXpert; Acc s+: 81.27. - [Fair Classification on CelebA (clean)](): Fair Classification; CelebA (clean); Accuracy (s+): 85.74. - [Fair Classification on CelebA clean (test)](): Fair Classification; CelebA clean (test); Accuracy s+: 97.34. - [Fair Classification on CelebA level-3 snow noise (corrupted)](): Fair Classification; CelebA level-3 snow noise (corrupted); Acc s+: 81.75. - [Fair Classification on CelebA corrupted level-3 Gaussian noise (test)](): Fair Classification; CelebA corrupted level-3 Gaussian noise (test); Accuracy (s+): 95.32. - [Gender Classification on FairFace (test)](): Gender Classification; FairFace (test); Overall Accuracy: 83.94. ## Original sources and implementation [Original paper]() [PDF]()