Towards Fairness under Label Bias in Image Segmentation: Impact, Measurement and Mitigation

2026-05-07

AI Summary

This work addresses the challenge of label bias in image segmentation, where annotation pipelines induce systematic performance disparities across demographic subgroups—a problem exacerbated by the absence of unbiased ground-truth labels for detection and mitigation. The study presents the first adaptation of Confident Learning to image segmentation, leveraging directional errors between model predictions and training labels to identify and quantify label bias without requiring clean annotations. It further reveals that such bias manifests as subgroup separability in the encoder’s feature space and proposes a novel debiasing strategy that operates without access to unbiased labels. Experiments on three datasets—encompassing both synthetic and real-world label bias—demonstrate that the method reliably detects bias and significantly improves segmentation fairness across subgroups.

Abstract

Labeled datasets reflect the biases of their annotation pipelines, which sometimes introduce label bias: group-conditional label errors that cause systematic performance disparities across demographic subgroups. Label bias in image segmentation remains underexplored, as even detecting it typically requires clean, unbiased annotations, which are not readily available. We present a data-centric adaptation of Confident Learning to segmentation, allowing detection of label bias directly in the training data without a clean, unbiased ground truth. By comparing the provided training labels to the model's confident predictions, we isolate directional errors that quantify the presence and nature of bias, where standard overlap metrics like Dice fail. We further show that label bias influences subgroup separability in the encoder's feature space, an artifact we leverage for bias mitigation rather than suppressing it. We evaluate three datasets, spanning from synthetic to real-life bias, showing how our framework reliably detects and mitigates bias without access to clean labels, achieving equitable performance across experimental conditions.

Problem

• label bias
• image segmentation
• fairness
• demographic subgroups
• annotation bias

Innovation

• Label Bias
• Confident Learning
• Image Segmentation
• Fairness
• Bias Mitigation