Mitigating Context Bias in Domain Adaptation for Object Detection using Mask Pooling

2025-05-24

AI Summary

This work addresses context bias in domain adaptive object detection (DAOD), arising from spurious foreground-background correlations. For the first time, it identifies convolutional pooling operations as the causal source of this bias through a causal inference lens. To mitigate it, we propose Mask Pooling: a mechanism that decouples foreground and background pooling pathways via foreground masks, thereby suppressing background dependency; and a fully randomized background benchmark to rigorously evaluate domain robustness. Integrated into mainstream frameworks (e.g., Mask R-CNN), our approach combines causal modeling with domain-adaptive training. Experiments on multiple DAOD benchmarks show consistent mAP gains of 3.2–5.8 points. Moreover, the method demonstrates remarkable stability under extreme background perturbations, significantly improving invariance to contextual interference while preserving detection accuracy.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Context bias refers to the association between the foreground objects and background during the object detection training process. Various methods have been proposed to minimize the context bias when applying the trained model to an unseen domain, known as domain adaptation for object detection (DAOD). But a principled approach to understand why the context bias occurs and how to remove it has been missing. In this work, we provide a causal view of the context bias, pointing towards the pooling operation in the convolution network architecture as the possible source of this bias. We present an alternative, Mask Pooling, which uses an additional input of foreground masks, to separate the pooling process in the respective foreground and background regions and show that this process leads the trained model to detect objects in a more robust manner under different domains. We also provide a benchmark designed to create an ultimate test for DAOD, using foregrounds in the presence of absolute random backgrounds, to analyze the robustness of the intended trained models. Through these experiments, we hope to provide a principled approach for minimizing context bias under domain shift.

Problem

• Understanding and mitigating context bias in domain adaptation for object detection
• Proposing Mask Pooling to separate foreground and background pooling processes
• Creating a benchmark to test robustness of models against random backgrounds

Innovation

• Introduces Mask Pooling to separate foreground and background
• Uses foreground masks for robust object detection
• Provides a benchmark with random backgrounds for testing