Unbiased Faster R-CNN for Single-source Domain Generalized Object Detection

TL;DR

This paper introduces Unbiased Faster R-CNN, a novel approach for single-source domain generalization in object detection that leverages causal modeling and data augmentation to improve robustness across diverse scenes.

Contribution

The paper proposes a causal perspective for SDG, designing modules for data augmentation, attention learning, and prototype learning to mitigate data and feature biases.

Findings

Achieves 3.9% mAP improvement on Night-Clear scene.

Effectively mitigates data bias through data augmentation.

Enhances generalization ability across five scenes.

Abstract

Single-source domain generalization (SDG) for object detection is a challenging yet essential task as the distribution bias of the unseen domain degrades the algorithm performance significantly. However, existing methods attempt to extract domain-invariant features, neglecting that the biased data leads the network to learn biased features that are non-causal and poorly generalizable. To this end, we propose an Unbiased Faster R-CNN (UFR) for generalizable feature learning. Specifically, we formulate SDG in object detection from a causal perspective and construct a Structural Causal Model (SCM) to analyze the data bias and feature bias in the task, which are caused by scene confounders and object attribute confounders. Based on the SCM, we design a Global-Local Transformation module for data augmentation, which effectively simulates domain diversity and mitigates the data bias. Additionally, we introduce a Causal Attention Learning module that incorporates a designed attention invariance loss to learn image-level features that are robust to scene confounders. Moreover, we develop a Causal Prototype Learning module with an explicit instance constraint and an implicit prototype constraint, which further alleviates the negative impact of object attribute confounders. Experimental results on five scenes demonstrate the prominent generalization ability of our method, with an improvement of 3.9% mAP on the Night-Clear scene.