Leveraging Confident Image Regions for Source-Free Domain-Adaptive Object Detection

TL;DR

This paper introduces a novel data augmentation method for source-free domain-adaptive object detection that leverages confident target image regions, improving adaptation without source data and achieving state-of-the-art results.

Contribution

It proposes a new data augmentation scheme tailored for source-free domain adaptation, using confident region cutouts and a teacher-student framework to enhance detector adaptation.

Findings

Achieved state-of-the-art results on two traffic scene benchmarks.

Introduced a confident region-based augmentation technique.

Demonstrated effectiveness of teacher-student paradigm in source-free setting.

Abstract

Source-free domain-adaptive object detection is an interesting but scarcely addressed topic. It aims at adapting a source-pretrained detector to a distinct target domain without resorting to source data during adaptation. So far, there is no data augmentation scheme tailored to source-free domain-adaptive object detection. To this end, this paper presents a novel data augmentation approach that cuts out target image regions where the detector is confident, augments them along with their respective pseudo-labels, and joins them into a challenging target image to adapt the detector. As the source data is out of reach during adaptation, we implement our approach within a teacher-student learning paradigm to ensure that the model does not collapse during the adaptation procedure. We evaluated our approach on three adaptation benchmarks of traffic scenes, scoring new state-of-the-art on two of them.