Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

EReLiFM: Evidential Reliability-Aware Residual Flow Meta-Learning for Open-Set Domain Generalization under Noisy Labels

Kunyu Peng, Di Wen, Kailun Yang, Jia Fu, Yufan Chen, Ruiping Liu, Jiamin Wu, Junwei Zheng, M. Saquib Sarfraz, Luc Van Gool, Danda Pani Paudel, Rainer Stiefelhagen

TL;DR

Open-Set Domain Generalization under Noisy Labels remains difficult due to concurrent domain shift, unseen categories, and label corruption. EReLiFM combines reliability-aware evidential training dynamics for clean/noisy data separation with Domain and Category Conditioned Residual Flow Matching to model diverse, uncertainty-aware transfers across domains and categories within a meta-learning loop. Across PACS, DigitsDG, and TerraIncBeery, it achieves state-of-the-art results under symmetric and asymmetric noise, highlighting robust noise diagnosis, data expansion through structured residuals, and improved generalization to unseen domains. The approach offers practical impact by enabling more trustworthy open-set recognition in real-world data with noisy annotations.

Abstract

Open-Set Domain Generalization (OSDG) aims to enable deep learning models to recognize unseen categories in new domains, which is crucial for real-world applications. Label noise hinders open-set domain generalization by corrupting source-domain knowledge, making it harder to recognize known classes and reject unseen ones. While existing methods address OSDG under Noisy Labels (OSDG-NL) using hyperbolic prototype-guided meta-learning, they struggle to bridge domain gaps, especially with limited clean labeled data. In this paper, we propose Evidential Reliability-Aware Residual Flow Meta-Learning (EReLiFM). We first introduce an unsupervised two-stage evidential loss clustering method to promote label reliability awareness. Then, we propose a residual flow matching mechanism that models structured domain- and category-conditioned residuals, enabling diverse and uncertainty-aware transfer paths beyond interpolation-based augmentation. During this meta-learning process, the model is optimized such that the update direction on the clean set maximizes the loss decrease on the noisy set, using pseudo labels derived from the most confident predicted class for supervision. Experimental results show that EReLiFM outperforms existing methods on OSDG-NL, achieving state-of-the-art performance. The source code is available at https://github.com/KPeng9510/ERELIFM.

EReLiFM: Evidential Reliability-Aware Residual Flow Meta-Learning for Open-Set Domain Generalization under Noisy Labels

TL;DR

Open-Set Domain Generalization under Noisy Labels remains difficult due to concurrent domain shift, unseen categories, and label corruption. EReLiFM combines reliability-aware evidential training dynamics for clean/noisy data separation with Domain and Category Conditioned Residual Flow Matching to model diverse, uncertainty-aware transfers across domains and categories within a meta-learning loop. Across PACS, DigitsDG, and TerraIncBeery, it achieves state-of-the-art results under symmetric and asymmetric noise, highlighting robust noise diagnosis, data expansion through structured residuals, and improved generalization to unseen domains. The approach offers practical impact by enabling more trustworthy open-set recognition in real-world data with noisy annotations.

Abstract

Open-Set Domain Generalization (OSDG) aims to enable deep learning models to recognize unseen categories in new domains, which is crucial for real-world applications. Label noise hinders open-set domain generalization by corrupting source-domain knowledge, making it harder to recognize known classes and reject unseen ones. While existing methods address OSDG under Noisy Labels (OSDG-NL) using hyperbolic prototype-guided meta-learning, they struggle to bridge domain gaps, especially with limited clean labeled data. In this paper, we propose Evidential Reliability-Aware Residual Flow Meta-Learning (EReLiFM). We first introduce an unsupervised two-stage evidential loss clustering method to promote label reliability awareness. Then, we propose a residual flow matching mechanism that models structured domain- and category-conditioned residuals, enabling diverse and uncertainty-aware transfer paths beyond interpolation-based augmentation. During this meta-learning process, the model is optimized such that the update direction on the clean set maximizes the loss decrease on the noisy set, using pseudo labels derived from the most confident predicted class for supervision. Experimental results show that EReLiFM outperforms existing methods on OSDG-NL, achieving state-of-the-art performance. The source code is available at https://github.com/KPeng9510/ERELIFM.
Paper Structure (23 sections, 7 equations, 1 figure, 18 tables, 1 algorithm)

This paper contains 23 sections, 7 equations, 1 figure, 18 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Task Description
  5. EReLiFM
  6. Experiments
  7. Noisy Label Settings
  8. Datasets and Metrics
  9. Implementation Details
  10. Analysis of the Model Performance
  11. Cross-Backbone Generalizability
  12. Evaluation on Another Dataset
  13. Analysis of the Module Ablation
  14. Conclusion
  15. The Use of Large Language Models (LLMs)
  16. ...and 8 more sections

Figures (1)

  • Figure 1: Confidence score visualization of learned representations on PACS with target domain photo, using ResNet18 he2016deep under symmetric label noise with ratio $50\%$.