Few-Shot Object Detection in Unseen Domains

TL;DR

This paper introduces a novel approach for few-shot object detection across unseen domains, combining meta-training and data augmentation to learn domain-agnostic representations without target domain labels.

Contribution

It proposes a two-fold method involving meta-training for domain shift and data augmentation, along with a contrastive loss to improve domain generalization in FSOD.

Findings

Significantly reduces domain gap in FSOD tasks

Effective on T-LESS, PASCAL-VOC, and ExDark datasets

No need for target domain labels or images of novel classes

Abstract

Few-shot object detection (FSOD) has thrived in recent years to learn novel object classes with limited data by transferring knowledge gained on abundant base classes. FSOD approaches commonly assume that both the scarcely provided examples of novel classes and test-time data belong to the same domain. However, this assumption does not hold in various industrial and robotics applications, where a model can learn novel classes from a source domain while inferring on classes from a target domain. In this work, we address the task of zero-shot domain adaptation, also known as domain generalization, for FSOD. Specifically, we assume that neither images nor labels of the novel classes in the target domain are available during training. Our approach for solving the domain gap is two-fold. First, we leverage a meta-training paradigm, where we learn the domain shift on the base classes, then transfer the domain knowledge to the novel classes. Second, we propose various data augmentations techniques on the few shots of novel classes to account for all possible domain-specific information. To constraint the network into encoding domain-agnostic class-specific representations only, a contrastive loss is proposed to maximize the mutual information between foreground proposals and class embeddings and reduce the network's bias to the background information from target domain. Our experiments on the T-LESS, PASCAL-VOC, and ExDark datasets show that the proposed approach succeeds in alleviating the domain gap considerably without utilizing labels or images of novel categories from the target domain.