Decoupled Adaptation for Cross-Domain Object Detection

TL;DR

This paper introduces D-adapt, a decoupled adaptation method for cross-domain object detection that improves transferability by separately handling feature and bounding box regression adaptation, achieving state-of-the-art results.

Contribution

The paper proposes a novel decoupled adaptation framework that separately optimizes feature and bounding box regression adaptation for cross-domain object detection.

Findings

Achieves state-of-the-art results on four cross-domain detection tasks.

Yields 17% and 21% relative improvements on Clipart1k and Comic2k datasets.

Effectively separates feature and regression adaptation processes.

Abstract

Cross-domain object detection is more challenging than object classification since multiple objects exist in an image and the location of each object is unknown in the unlabeled target domain. As a result, when we adapt features of different objects to enhance the transferability of the detector, the features of the foreground and the background are easy to be confused, which may hurt the discriminability of the detector. Besides, previous methods focused on category adaptation but ignored another important part for object detection, i.e., the adaptation on bounding box regression. To this end, we propose D-adapt, namely Decoupled Adaptation, to decouple the adversarial adaptation and the training of the detector. Besides, we fill the blank of regression domain adaptation in object detection by introducing a bounding box adaptor. Experiments show that D-adapt achieves state-of-the-art results on four cross-domain object detection tasks and yields 17% and 21% relative improvement on benchmark datasets Clipart1k and Comic2k in particular.