Search papers, labs, and topics across Lattice.
This paper addresses the challenge of Single-Domain Generalized Object Detection (Single-DGOD) by proposing a novel approach called Manifold Regression with Visual-Text Dual Chain-of-Thought (MR-DCoT) that rectifies deviant samples back to a semantic manifold rather than relying on traditional simulation-driven strategies. The authors argue that existing methods often lead to overfitting and limited robustness due to their dependence on synthetic data, which fails to capture the complexities of real-world scenarios. Through extensive experiments across various benchmarks, MR-DCoT shows significant improvements in generalization and robustness against unseen domain shifts, outperforming conventional techniques.
Rectifying off-manifold samples to a stable semantic manifold can dramatically enhance object detection performance in unseen domains.
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.
