Data-Centric Benchmark for Label Noise Estimation and Ranking in Remote Sensing Image Segmentation

2026-02-28

AI Summary

This work addresses the challenges of high annotation costs and label noise in remote sensing image semantic segmentation, which severely degrade model performance and robustness. To this end, we present the first benchmark specifically designed for evaluating and ranking label noise in remote sensing segmentation tasks. We propose two complementary noise quantification methods that jointly leverage model uncertainty, prediction consistency, and feature representation analysis to holistically assess the noise level of training samples. Alongside this benchmark, we release the first publicly available remote sensing dataset with annotated label noise. Extensive experiments under diverse settings demonstrate that our approach significantly outperforms existing baselines. The code and dataset are made openly accessible to facilitate future research.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

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: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

High-quality pixel-level annotations are essential for the semantic segmentation of remote sensing imagery. However, such labels are expensive to obtain and often affected by noise due to the labor-intensive and time-consuming nature of pixel-wise annotation, which makes it challenging for human annotators to label every pixel accurately. Annotation errors can significantly degrade the performance and robustness of modern segmentation models, motivating the need for reliable mechanisms to identify and quantify noisy training samples. This paper introduces a novel Data-Centric benchmark, together with a novel, publicly available dataset and two techniques for identifying, quantifying, and ranking training samples according to their level of label noise in remote sensing semantic segmentation. Such proposed methods leverage complementary strategies based on model uncertainty, prediction consistency, and representation analysis, and consistently outperform established baselines across a range of experimental settings. The outcomes of this work are publicly available at https://github.com/keillernogueira/label_noise_segmentation.

Problem

• label noise
• remote sensing
• semantic segmentation
• data quality
• annotation error

Innovation

• label noise estimation
• remote sensing segmentation
• data-centric benchmark
• model uncertainty
• prediction consistency