Unsupervised Domain Adaptation for Segmentation with Black-box Source Model

TL;DR

This paper proposes a novel unsupervised domain adaptation method for segmentation that uses only a black-box source model, employing knowledge distillation and entropy minimization to adapt to the target domain without access to source data.

Contribution

It introduces a practical UDA approach relying solely on a black-box source model, avoiding the need for source data or white-box access, with a new knowledge distillation scheme.

Findings

Achieved performance comparable to white-box source model adaptation.

Validated on BraTS 2018 database.

Effective in privacy-preserving cross-domain segmentation.

Abstract

Unsupervised domain adaptation (UDA) has been widely used to transfer knowledge from a labeled source domain to an unlabeled target domain to counter the difficulty of labeling in a new domain. The training of conventional solutions usually relies on the existence of both source and target domain data. However, privacy of the large-scale and well-labeled data in the source domain and trained model parameters can become the major concern of cross center/domain collaborations. In this work, to address this, we propose a practical solution to UDA for segmentation with a black-box segmentation model trained in the source domain only, rather than original source data or a white-box source model. Specifically, we resort to a knowledge distillation scheme with exponential mixup decay (EMD) to gradually learn target-specific representations. In addition, unsupervised entropy minimization is further applied to regularization of the target domain confidence. We evaluated our framework on the BraTS 2018 database, achieving performance on par with white-box source model adaptation approaches.