Cross-Domain Weakly-Supervised Object Detection through Progressive Domain Adaptation

TL;DR

This paper introduces a novel cross-domain weakly supervised object detection framework that adapts a source domain detector to target domains with only image-level annotations, achieving significant performance improvements.

Contribution

It proposes a two-step progressive domain adaptation method that fine-tunes a pre-trained detector using artificially generated samples for cross-domain weakly supervised detection.

Findings

Achieved 5-20% mAP improvement over baselines.

Validated on newly collected multi-domain datasets.

Demonstrated effectiveness of progressive adaptation approach.

Abstract

Can we detect common objects in a variety of image domains without instance-level annotations? In this paper, we present a framework for a novel task, cross-domain weakly supervised object detection, which addresses this question. For this paper, we have access to images with instance-level annotations in a source domain (e.g., natural image) and images with image-level annotations in a target domain (e.g., watercolor). In addition, the classes to be detected in the target domain are all or a subset of those in the source domain. Starting from a fully supervised object detector, which is pre-trained on the source domain, we propose a two-step progressive domain adaptation technique by fine-tuning the detector on two types of artificially and automatically generated samples. We test our methods on our newly collected datasets containing three image domains, and achieve an improvement of approximately 5 to 20 percentage points in terms of mean average precision (mAP) compared to the best-performing baselines.