Learnable Data Augmentation for One-Shot Unsupervised Domain Adaptation

TL;DR

This paper introduces LearnAug-UDA, a learnable data augmentation framework for one-shot unsupervised domain adaptation, which enhances source data to resemble the target with only a single unlabeled target sample, improving model generalization.

Contribution

It proposes a novel encoder-decoder based augmentation method that uses perceptual loss and style transfer for effective domain adaptation with minimal target data.

Findings

Achieves state-of-the-art results on DomainNet and VisDA benchmarks.

Effectively augments source data to match target domain characteristics.

Demonstrates robustness with only one target sample available.

Abstract

This paper presents a classification framework based on learnable data augmentation to tackle the One-Shot Unsupervised Domain Adaptation (OS-UDA) problem. OS-UDA is the most challenging setting in Domain Adaptation, as only one single unlabeled target sample is assumed to be available for model adaptation. Driven by such single sample, our method LearnAug-UDA learns how to augment source data, making it perceptually similar to the target. As a result, a classifier trained on such augmented data will generalize well for the target domain. To achieve this, we designed an encoder-decoder architecture that exploits a perceptual loss and style transfer strategies to augment the source data. Our method achieves state-of-the-art performance on two well-known Domain Adaptation benchmarks, DomainNet and VisDA. The project code is available at https://github.com/IIT-PAVIS/LearnAug-UDA