Quantifying Context Bias in Domain Adaptation for Object Detection

2024-09-23

arXiv.org

AI Summary

This work addresses foreground-background contextual bias induced by background context in domain adaptive object detection (DAOD), a previously unquantified challenge. We propose the first layer-specific contextual bias quantification framework based on conditional probability estimation, systematically characterizing cross-domain background feature misalignment. Methodologically, we integrate hierarchical activation modulation with learnable background mask intervention to enable interpretable, targeted suppression of background bias. Evaluated on CARLA, Cityscapes, and Foggy Cityscapes, our framework significantly mitigates performance degradation under adverse conditions such as fog, yielding a +3.2% mAP improvement. Key contributions include: (1) the first quantifiable metric for contextual bias in DAOD; (2) a novel bias-aware hierarchical adaptation paradigm; and (3) an interpretable solution that jointly supports diagnostic analysis and optimization.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Adversarial Attacks & Robustness
• Search and Optimization: Adversarial Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Systems and Infrastructure for Web, Mobile and WoT: Location- and context-aware Web and WoT applications and services

Abstract

Domain adaptation for object detection (DAOD) aims to transfer a trained model from a source to a target domain. Various DAOD methods exist, some of which minimize context bias between foreground-background associations in various domains. However, no prior work has studied context bias in DAOD by analyzing changes in background features during adaptation and how context bias is represented in different domains. Our research experiment highlights the potential usability of context bias in DAOD. We address the problem by varying activation values over different layers of trained models and by masking the background, both of which impact the number and quality of detections. We then use one synthetic dataset from CARLA and two different versions of real open-source data, Cityscapes and Cityscapes foggy, as separate domains to represent and quantify context bias. We utilize different metrics such as Maximum Mean Discrepancy (MMD) and Maximum Variance Discrepancy (MVD) to find the layer-specific conditional probability estimates of foreground given manipulated background regions for separate domains. We demonstrate through detailed analysis that understanding of the context bias can affect DAOD approach and foc

Problem

• Quantifying context bias in domain adaptation for object detection
• Analyzing background feature changes and context bias representation across domains
• Improving DAOD accuracy by addressing foreground-background association biases

Innovation

• Analyzes background feature changes in DAOD
• Uses MMD and MVD for conditional probability estimates
• Alleviates context bias improving detection accuracy