English

On Direct Distribution Matching for Adapting Segmentation Networks

Computer Vision and Pattern Recognition 2021-11-29 v2

Abstract

Minimization of distribution matching losses is a principled approach to domain adaptation in the context of image classification. However, it is largely overlooked in adapting segmentation networks, which is currently dominated by adversarial models. We propose a class of loss functions, which encourage direct kernel density matching in the network-output space, up to some geometric transformations computed from unlabeled inputs. Rather than using an intermediate domain discriminator, our direct approach unifies distribution matching and segmentation in a single loss. Therefore, it simplifies segmentation adaptation by avoiding extra adversarial steps, while improving both the quality, stability and efficiency of training. We juxtapose our approach to state-of-the-art segmentation adaptation via adversarial training in the network-output space. In the challenging task of adapting brain segmentation across different magnetic resonance images (MRI) modalities, our approach achieves significantly better results both in terms of accuracy and stability.

Keywords

Cite

@article{arxiv.1904.02657,
  title  = {On Direct Distribution Matching for Adapting Segmentation Networks},
  author = {Georg Pichler and Jose Dolz and Ismail Ben Ayed and Pablo Piantanida},
  journal= {arXiv preprint arXiv:1904.02657},
  year   = {2021}
}

Comments

includes appendix; published at MIDL2020: https://2020.midl.io/papers/pichler20.html

R2 v1 2026-06-23T08:29:32.653Z