Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PhysAug: A Physical-guided and Frequency-based Data Augmentation for Single-Domain Generalized Object Detection

Xiaoran Xu, Jiangang Yang, Wenhui Shi, Siyuan Ding, Luqing Luo, Jian Liu

TL;DR

PhysAug, a novel physical model-based non-ideal imaging condition data augmentation method, is proposed to enhance the adaptability of the S-DGOD tasks and significantly outperforms the state-of-the-art across various S-DGOD datasets.

Abstract

Single-Domain Generalized Object Detection~(S-DGOD) aims to train on a single source domain for robust performance across a variety of unseen target domains by taking advantage of an object detector. Existing S-DGOD approaches often rely on data augmentation strategies, including a composition of visual transformations, to enhance the detector's generalization ability. However, the absence of real-world prior knowledge hinders data augmentation from contributing to the diversity of training data distributions. To address this issue, we propose PhysAug, a novel physical model-based non-ideal imaging condition data augmentation method, to enhance the adaptability of the S-DGOD tasks. Drawing upon the principles of atmospheric optics, we develop a universal perturbation model that serves as the foundation for our proposed PhysAug. Given that visual perturbations typically arise from the interaction of light with atmospheric particles, the image frequency spectrum is harnessed to simulate real-world variations during training. This approach fosters the detector to learn domain-invariant representations, thereby enhancing its ability to generalize across various settings. Without altering the network architecture or loss function, our approach significantly outperforms the state-of-the-art across various S-DGOD datasets. In particular, it achieves a substantial improvement of $7.3\%$ and $7.2\%$ over the baseline on DWD and Cityscape-C, highlighting its enhanced generalizability in real-world settings.

PhysAug: A Physical-guided and Frequency-based Data Augmentation for Single-Domain Generalized Object Detection

TL;DR

PhysAug, a novel physical model-based non-ideal imaging condition data augmentation method, is proposed to enhance the adaptability of the S-DGOD tasks and significantly outperforms the state-of-the-art across various S-DGOD datasets.

Abstract

Single-Domain Generalized Object Detection~(S-DGOD) aims to train on a single source domain for robust performance across a variety of unseen target domains by taking advantage of an object detector. Existing S-DGOD approaches often rely on data augmentation strategies, including a composition of visual transformations, to enhance the detector's generalization ability. However, the absence of real-world prior knowledge hinders data augmentation from contributing to the diversity of training data distributions. To address this issue, we propose PhysAug, a novel physical model-based non-ideal imaging condition data augmentation method, to enhance the adaptability of the S-DGOD tasks. Drawing upon the principles of atmospheric optics, we develop a universal perturbation model that serves as the foundation for our proposed PhysAug. Given that visual perturbations typically arise from the interaction of light with atmospheric particles, the image frequency spectrum is harnessed to simulate real-world variations during training. This approach fosters the detector to learn domain-invariant representations, thereby enhancing its ability to generalize across various settings. Without altering the network architecture or loss function, our approach significantly outperforms the state-of-the-art across various S-DGOD datasets. In particular, it achieves a substantial improvement of and over the baseline on DWD and Cityscape-C, highlighting its enhanced generalizability in real-world settings.

Paper Structure

This paper contains 27 sections, 9 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Single-Domain Generalization for Object Detection
  4. Data Augmentation for Domain Generalization
  5. Preliminary
  6. Problem Formulation
  7. Evaluation Metrics
  8. Methodology
  9. A Physical-based Perturbation Model
  10. The Design of PhysAug
  11. Overview
  12. Global Non-uniform Illumination
  13. Partical-induced Local Occlusion
  14. Experiments
  15. Datasets
  16. ...and 12 more sections

Figures (4)

  • Figure 1: (a) The Physical imaging process in the atmosphere. Path1: Incident light path, Path2: Reflected light path, Path3: Atmospheric light path. (b) The cases of image degradation due to the different interactions between light and atmospheric particles.
  • Figure 2: An overview of PhysAug. The input of PhysAug is an RGB training image, and its output is an augmented image. The two important components in PhysAug are Global Non-uniform Illumination (Top) and Particle-induced Local Occlusion (Bottom).
  • Figure 3: Visualized detection samples of the baseline method, OA-Mix and PhysAug in different weather conditions.
  • Figure 4: Heatmaps comparison in a night-rainy scenario. Top-left: original image. Top-right: The baseline method. Bottom-left: OA-Mix. Bottom-right: PhysAug.