DG-DETR: Toward Domain Generalized Detection Transformer

2025-04-28

AI Summary

This paper addresses the limited out-of-distribution (OOD) robustness of end-to-end detection Transformers (DETRs) in domain generalization (DG) settings. To this end, we propose a plug-and-play DG framework for DETR without modifying the backbone or training pipeline. Our method introduces two key components: (1) a domain-agnostic query selection mechanism to mitigate cross-domain query bias; and (2) a wavelet-based feature disentanglement module that separates semantic content from style, followed by orthogonal projection and instance-level style space modeling to synthesize diverse, domain-invariant representations. Evaluated on multiple cross-domain object detection benchmarks—including PACS-D and DomainNet-D—our approach consistently improves OOD detection performance, achieving average mAP gains of 3.2–5.8 points over baseline DETR. The framework demonstrates strong generalizability and deployment flexibility, requiring no architectural or optimization modifications to the underlying detector.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Adversarial Attacks & Robustness
• Application Domains: Security

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies

Abstract

End-to-end Transformer-based detectors (DETRs) have demonstrated strong detection performance. However, domain generalization (DG) research has primarily focused on convolutional neural network (CNN)-based detectors, while paying little attention to enhancing the robustness of DETRs. In this letter, we introduce a Domain Generalized DEtection TRansformer (DG-DETR), a simple, effective, and plug-and-play method that improves out-of-distribution (OOD) robustness for DETRs. Specifically, we propose a novel domain-agnostic query selection strategy that removes domain-induced biases from object queries via orthogonal projection onto the instance-specific style space. Additionally, we leverage a wavelet decomposition to disentangle features into domain-invariant and domain-specific components, enabling synthesis of diverse latent styles while preserving the semantic features of objects. Experimental results validate the effectiveness of DG-DETR. Our code is available at https://github.com/sminhwang/DG-DETR.

Problem

• Enhancing DETRs' robustness for domain generalization
• Removing domain biases via orthogonal query selection
• Disentangling features into domain-invariant and specific components

Innovation

• Domain-agnostic query selection via orthogonal projection
• Wavelet decomposition for disentangling domain-invariant features
• Plug-and-play method enhancing DETR's OOD robustness