Prototype-Anchored Generalized Manifold Regression for Unknown-Domain Object Detection

2026-07-08

AI Summary

This work addresses the degraded generalization performance of detectors trained on a single source domain when deployed in unknown target domains, where structural degradation and stylistic shifts commonly occur. To tackle this challenge, the authors propose a manifold regression–based domain generalization framework that formulates out-of-domain generalization as a regression task on a semantic manifold. The method generates hard out-of-manifold samples through visual–textual dual-chain guided semantic evolution and diffusion-model-driven structural perturbations, then leverages a class-specific prototype anchoring mechanism to regress their features back onto the source semantic manifold. This approach jointly enforces semantic consistency and structural robustness without exhaustively enumerating perturbations. Extensive experiments demonstrate that the proposed method significantly outperforms existing approaches across three challenging benchmarks: adverse-weather object detection, real-to-artistic style generalization, and zero-shot semantic segmentation, confirming its effectiveness and broad applicability.

Abstract

In this paper, we study Single-Domain Generalized Object Detection (Single-DGOD), which aims to transfer a detector trained on a single source domain to multiple unseen domains. Existing methods mainly rely on simulation-driven strategies, such as data augmentation or textual prompts, to enlarge the training distribution. However, finite simulations can hardly cover the dynamic variations of real-world scenarios, often causing overfitting to synthetic styles and limited robustness to complex structural degradations. Inspired by the manifold hypothesis, we argue that semantic features, despite diverse visual changes, should lie on a compact and stable low-dimensional manifold. Therefore, robust generalization requires rectifying deviant samples back to this semantic manifold, rather than exhaustively simulating external perturbations. To this end, we propose Manifold Regression with Visual-Text Dual Chain-of-Thought (MR-DCoT), which formulates unknown-domain generalization as a manifold regression problem. MR-DCoT first uses a Visual-Text Dual Chain-of-Thought module to combine VLM-guided semantic evolution with diffusion-based structural perturbation, generating structured off-manifold hard examples. It then introduces Class-Specific Prototype Anchoring to learn a rectification operator that projects deviant features toward the source semantic manifold. By integrating outlier generation and semantic correction into a closed loop, MR-DCoT effectively narrows the distribution gap and improves robustness under unseen shifts. Extensive experiments on three complementary benchmarks, including adverse-weather detection, real-to-art generalization, and zero-shot semantic segmentation, demonstrate the effectiveness and versatility of our method.

Problem

• Single-Domain Generalized Object Detection
• unknown-domain generalization
• semantic manifold
• distribution shift
• object detection

Innovation

• Manifold Regression
• Prototype Anchoring
• Visual-Text Dual Chain-of-Thought
• Single-Domain Generalized Object Detection
• Out-of-Distribution Generalization