Rethinking Weak-to-Strong Augmentation in Source-Free Domain Adaptive Object Detection

2024-10-07

arXiv.org

AI Summary

Strong data augmentation in mean teacher-based source-free domain adaptive object detection often erases category semantics, leading to noisy pseudo-labels and class confusion. This work first identifies the critical semantic degradation problem arising from weak–strong augmentation pairs. To address it, we propose Weak–Strong Contrastive Learning (WSCoL): a dual-branch architecture with a mapping network ensures feature-space alignment; weak-feature-guided lossless knowledge distillation preserves semantic fidelity; and prototype-guided adaptive cross-class contrastive learning—augmented by uncertainty-aware background contrast optimization—enhances discriminability. Evaluated on multiple source-free object detection (SFOD) benchmarks, WSCoL achieves state-of-the-art performance, significantly improving cross-domain robustness and generalization. Crucially, it establishes an intrinsic semantic preservation mechanism for the mean teacher paradigm, overcoming a fundamental limitation of prior augmentation-dependent approaches.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Adversarial Attacks & Robustness
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Source-Free domain adaptive Object Detection (SFOD) aims to transfer a detector (pre-trained on source domain) to new unlabelled target domains. Current SFOD methods typically follow the Mean Teacher framework, where weak-to-strong augmentation provides diverse and sharp contrast for self-supervised learning. However, this augmentation strategy suffers from an inherent problem called crucial semantics loss: Due to random, strong disturbance, strong augmentation is prone to losing typical visual components, hindering cross-domain feature extraction. To address this thus-far ignored limitation, this paper introduces a novel Weak-to-Strong Contrastive Learning (WSCoL) approach. The core idea is to distill semantics lossless knowledge in the weak features (from the weak/teacher branch) to guide the representation learning upon the strong features (from the strong/student branch). To achieve this, we project the original features into a shared space using a mapping network, thereby reducing the bias between the weak and strong features. Meanwhile, a weak features-guided contrastive learning is performed in a weak-to-strong manner alternatively. Specifically, we first conduct an adaptation-aware prototype-guided clustering on the weak features to generate pseudo labels for corresponding strong features matched through proposals. Sequentially, we identify positive-negative samples based on the pseudo labels and perform cross-category contrastive learning on the strong features where an uncertainty estimator encourages adaptive background contrast. Extensive experiments demonstrate that WSCoL yields new state-of-the-art performance, offering a built-in mechanism mitigating crucial semantics loss for traditional Mean Teacher framework. The code and data will be released soon.

Problem

• Strong augmentation erases class-relevant semantic components
• Artificial inter-category confusion occurs in domain adaptation
• Compensating lost semantics in strongly augmented detection images

Innovation

• WSC compensates semantics lost in strong augmentation
• WSC uses weak augmentation as semantic anchors
• WSC is a plug-in for SFOD pipelines