Instance-Invariant Domain Adaptive Object Detection via Progressive Disentanglement

TL;DR

This paper introduces a progressive disentangled framework for domain adaptive object detection that extracts instance-invariant features by separating domain-invariant and domain-specific features, improving cross-domain detection performance.

Contribution

It proposes a novel progressive disentanglement approach with dedicated layers and a three-stage training mechanism for better domain-invariant feature extraction in object detection.

Findings

Outperforms baseline by 2.3%, 3.6%, and 4.0% on three domain-shift scenes.

Effectively disentangles domain-invariant features from domain-specific features.

Enhances generalization ability of object detectors across different domains.

Abstract

Most state-of-the-art methods of object detection suffer from poor generalization ability when the training and test data are from different domains, e.g., with different styles. To address this problem, previous methods mainly use holistic representations to align feature-level and pixel-level distributions of different domains, which may neglect the instance-level characteristics of objects in images. Besides, when transferring detection ability across different domains, it is important to obtain the instance-level features that are domain-invariant, instead of the styles that are domain-specific. Therefore, in order to extract instance-invariant features, we should disentangle the domain-invariant features from the domain-specific features. To this end, a progressive disentangled framework is first proposed to solve domain adaptive object detection. Particularly, base on disentangled learning used for feature decomposition, we devise two disentangled layers to decompose domain-invariant and domain-specific features. And the instance-invariant features are extracted based on the domain-invariant features. Finally, to enhance the disentanglement, a three-stage training mechanism including multiple loss functions is devised to optimize our model. In the experiment, we verify the effectiveness of our method on three domain-shift scenes. Our method is separately 2.3\%, 3.6\%, and 4.0\% higher than the baseline method \cite{saito2019strong}.