An assessment of data-centric methods for label noise identification in remote sensing data sets

Felix Kröber; Genc Hoxha; Ribana Roscher

cs.CVMar 17, 2026

An assessment of data-centric methods for label noise identification in remote sensing data sets

Authors: Felix Kröber, Genc Hoxha, Ribana Roscher

Organizations: Institute of Bio- and Geosciences, Forschungszentrum Jülich, Leo-Brandt-Straße, 52425 Jülich, Germany · Institute of Geodesy and Geoinformation, University of Bonn, Germany

Abstract

Label noise in the sense of incorrect labels is present in many real-world data sets and is known to severely limit the generalizability of deep learning models. In the field of remote sensing, however, automated treatment of label noise in data sets has received little attention to date. In particular, there is a lack of systematic analysis of the performance of data-centric methods that not only cope with label noise but also explicitly identify and isolate noisy labels. In this paper, we examine three such methods and evaluate their behavior under different label noise assumptions. To do this, we inject different types of label noise with noise levels ranging from 10 to 70% into two benchmark data sets, followed by an analysis of how well the selected methods filter the label noise and how this affects task performances. With our analyses, we clearly prove the value of data-centric methods for both parts - label noise identification and task performance improvements. Our analyses provide insights into which method is the best choice depending on the setting and objective. Finally, we show in which areas there is still a need for research in the transfer of data-centric label noise methods to remote sensing data. As such, our work is a step forward in bridging the methodological establishment of data-centric label noise methods and their usage in practical settings in the remote sensing domain.

Figures & tables

Figure 1: High-level overview of using data-centric methods to filter label noise and obtain a cleaned data set for model training.

Figure 2: Exemplary noise transition matrices for EuroSAT at an overall noise level of 50%.

Figure 3: Statistics on label noise identification across settings.

		TopoFilter	AuM	CL
Wins $\uparrow$	precision	12	7	5
	recall	13	11	0
	remaining noise	12	10	2
Losses $\uparrow$	precision	5	10	9
	recall	1	0	23
	remaining noise	1	1	22

Table 1: Summary of times a given data-centric method performs best or worst across all settings (4 noise levels, 3 noise types & 2 data sets) specified for the noise identification metrics.

Figure 4: Statistics on data set-wide label noise level estimation across settings.

		TopoFilter	AuM	CL
Wins $\uparrow$	$\Delta$ noise level	10	2	12
Wins $\uparrow$	SMAPE	10	3	11
Losses $\uparrow$	$\Delta$ noise level	9	12	3
Losses $\uparrow$	SMAPE	9	12	3

Table 2: Summary of times a given data-centric method performs best or worst across all settings (4 noise levels, 3 noise types & 2 data sets) specified for the noise level estimation metrics.

Figure 5: Scene classification test accuracies when training on cleaned and noisy data.

		TopoFilter	AuM	CL
Wins $\uparrow$	accuracy	11	9	4
Losses $\uparrow$	accuracy	9	8	7

Table 3: Summary of times a given data-centric method performs best or worst across all settings (4 noise levels, 3 noise types & 2 data sets) specified for the task performance metric.

Figure 6: UMAP-reduced representation of the label noise constellations examined. Input variables are shown in the left semicircle, the calculated label noise filtering properties in the right semicircle, and the final effect on task accuracies in the center. Task accuracy improvements are measured against the baseline of training on uncleaned data depicted as ”plain” in Figure 5 . Clusters with homogeneous behavior in terms of filtering properties and task accuracy improvements are marked with dashed lines (A-E).

Noisy-Label Learning

Noisy-label learning develops ways to train and evaluate models when some supervision is incorrect, ambiguous, or unreliable.

See all 93 papers

Data-Centric Benchmark for Label Noise Estimation and Ranking in Remote Sensing Binary Building Segmentation
Feb 28, 2026Keiller Nogueira, Codrut-Andrei Diaconu, Dávid Kerekes +9Learning with Noisy Labels Noisy-Label Learning
School of Computer Science and Informatics, University of Liverpool, Liverpool, L69 7ZX, UK · German Aerospace Center (DLR), Remote Sensing Technology Institute, Wessling, 82234, Germany · Division of Geoinformatics, KTH Royal Institute of Technology, Stockholm, 114 28, Sweden +4
The Re-Label Method For Data-Centric Machine Learning
Feb 9, 2023Tong GuoLearning with Noisy Labels Human-in-the-Loop Annotation
An Adaptive Data cleaning Framework for Noisy Label Detection
Jun 5, 2026Chen-Hsuan Fang, Wei-Hsinag Chen, Pin-Hsuan Yu +2Noisy-Label Learning Learning with Noisy Labels
Department of Electrical Eng. National Taiwan Ocean University, Keelung City 20224, Taiwan · AI Research Center, National Taiwan Ocean University, Keelung City 20224, Taiwan

Feb 28, 2026J/K move · Enter open · S save

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

Keiller Nogueira, Codrut-Andrei Diaconu, Dávid Kerekes, Jakob Gawlikowski+8

School of Computer Science and Informatics, University of Liverpool, Liverpool, L69 7ZX, UK · German Aerospace Center (DLR), Remote Sensing Technology Institute, Wessling, 82234, Germany · Division of Geoinformatics, KTH Royal Institute of Technology, Stockholm, 114 28, Sweden+4

Learning with Noisy Labels Noisy-Label Learning Remote Sensing Benchmarks Remote Sensing Image Segmentation

Open paper AskPDF arXiv

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 new, publicly available binary building segmentation dataset. This specific task serves as a representative and practically relevant testbed that enables controlled experimentation with different annotation perturbations. Furthermore, we also introduce 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.

Abstract

Noisy-Label Learning

Explore similar work

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

The Re-Label Method For Data-Centric Machine Learning

An Adaptive Data cleaning Framework for Noisy Label Detection