Segmentation with Noisy Labels via Spatially Correlated Distributions

2025-04-21

AI Summary

In semantic segmentation, remote sensing and medical imaging often suffer from spatially correlated annotation noise—such as mislabeling, missing annotations, and inter-annotator inconsistency—violating the standard assumption of independent label noise. This work is the first to model annotation noise as a spatially correlated stochastic process, leveraging a Gaussian random field prior with Kac–Murdock–Szegő (KMS)-structured covariance to enable scalable approximate Bayesian inference. The proposed framework explicitly captures noise dependencies among neighboring pixels, thereby enhancing model robustness to spatially structured label corruption. Extensive experiments across multiple tasks demonstrate substantial improvements over state-of-the-art noise-robust methods. Notably, on lung segmentation, the method achieves performance nearly matching that of fully clean-label training under moderate noise levels. These results validate both the effectiveness and practicality of explicitly modeling spatial correlation in annotation noise.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Semi-Supervised Learning
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Semantics and Knowledge: Representation, semantic annotation, enhancement, enrichments, access and/or integration of a variety of data on the Web
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

In semantic segmentation, the accuracy of models heavily depends on the high-quality annotations. However, in many practical scenarios such as medical imaging and remote sensing, obtaining true annotations is not straightforward and usually requires significant human labor. Relying on human labor often introduces annotation errors, including mislabeling, omissions, and inconsistency between annotators. In the case of remote sensing, differences in procurement time can lead to misaligned ground truth annotations. These label errors are not independently distributed, and instead usually appear in spatially connected regions where adjacent pixels are more likely to share the same errors. To address these issues, we propose an approximate Bayesian estimation based on a probabilistic model that assumes training data includes label errors, incorporating the tendency for these errors to occur with spatial correlations between adjacent pixels. Bayesian inference requires computing the posterior distribution of label errors, which becomes intractable when spatial correlations are present. We represent the correlation of label errors between adjacent pixels through a Gaussian distribution whose covariance is structured by a Kac-Murdock-Szeg""{o} (KMS) matrix, solving the computational challenges. Through experiments on multiple segmentation tasks, we confirm that leveraging the spatial correlation of label errors significantly improves performance. Notably, in specific tasks such as lung segmentation, the proposed method achieves performance comparable to training with clean labels under moderate noise levels. Code is available at https://github.com/pfnet-research/Bayesian_SpatialCorr.

Problem

• Addresses noisy labels in semantic segmentation tasks
• Models spatial correlation of label errors in adjacent pixels
• Improves segmentation accuracy under noisy annotation conditions

Innovation

• Bayesian estimation for noisy label segmentation
• Spatial correlation via KMS matrix
• Improves performance with correlated errors