Source-Free Domain Adaptive Object Detection with Semantics Compensation

TL;DR

This paper identifies that strong data augmentation in source-free domain adaptive object detection can erase class-relevant features, causing confusion, and proposes a semantics compensation method using weakly augmented images to improve detection accuracy.

Contribution

Introduces Weak-to-strong Semantics Compensation (WSCo), a plug-in method that mitigates semantic loss during augmentation, enhancing existing SFOD models' performance.

Findings

Strong augmentation can harm detection accuracy.

WSCo effectively restores semantics lost during augmentation.

Improved results on standard benchmarks.

Abstract

Strong data augmentation is a fundamental component of state-of-the-art mean teacher-based Source-Free domain adaptive Object Detection (SFOD) methods, enabling consistency-based self-supervised optimization along weak augmentation. However, our theoretical analysis and empirical observations reveal a critical limitation: strong augmentation can inadvertently erase class-relevant components, leading to artificial inter-category confusion. To address this issue, we introduce Weak-to-strong Semantics Compensation (WSCo), a novel remedy that leverages weakly augmented images, which preserve full semantics, as anchors to enrich the feature space of their strongly augmented counterparts. Essentially, this compensates for the class-relevant semantics that may be lost during strong augmentation on the fly. Notably, WSCo can be implemented as a generic plug-in, easily integrable with any existing SFOD pipelines. Extensive experiments validate the negative impact of strong augmentation on detection performance, and the effectiveness of WSCo in enhancing the performance of previous detection models on standard benchmarks.