D-GAP: Improving Out-of-Domain Robustness via Dataset-Agnostic and Gradient-Guided Augmentation in Frequency and Pixel Spaces

TL;DR

D-GAP is a novel augmentation method that enhances out-of-domain robustness in computer vision by adaptively blending frequency and pixel information based on model sensitivity, outperforming existing methods.

Contribution

The paper introduces D-GAP, a dataset-agnostic, gradient-guided augmentation technique in frequency and pixel spaces that improves out-of-domain robustness without requiring expert knowledge.

Findings

D-GAP improves OOD performance by +5.3% on real-world datasets.

D-GAP outperforms existing domain adaptation methods.

Extensive experiments validate the effectiveness of D-GAP.

Abstract

Out-of-domain (OOD) robustness is challenging to achieve in real-world computer vision applications, where shifts in image background, style, and acquisition instruments always degrade model performance. Generic augmentations show inconsistent gains under such shifts, whereas dataset-specific augmentations require expert knowledge and prior analysis. Moreover, prior studies show that neural networks adapt poorly to domain shifts because they exhibit a learning bias to domain-specific frequency components. Perturbing frequency values can mitigate such bias but overlooks pixel-level details, leading to suboptimal performance. To address these problems, we propose D-GAP, a Dataset-agnostic and Gradient-guided augmentation method for the Amplitude spectrum (in frequency space) and the Pixel values, improving OOD robustness by introducing targeted augmentation in both frequency and pixel spaces. Unlike conventional handcrafted augmentations, D-GAP computes sensitivity maps in the frequency space from task gradients, which reflect how strongly the deep models respond to different frequency components, and uses the maps to adaptively interpolate amplitudes between source and target samples. This way, D-GAP reduces the learning bias in frequency space, while a complementary pixel-space blending procedure restores fine spatial details. Extensive experiments on four real-world datasets and three domain-adaptation benchmarks show that D-GAP consistently outperforms both generic and dataset-specific domain adaptation methods, improving average OOD performance by +5.3% on real-world datasets and +1.9% on benchmark datasets.