Deep semi-supervised segmentation with weight-averaged consistency targets

TL;DR

This paper extends the Mean Teacher semi-supervised learning method to image segmentation, demonstrating improved performance on MRI data with small datasets and proposing solutions for data augmentation challenges.

Contribution

It adapts the Mean Teacher approach for segmentation tasks and introduces a method to address data augmentation issues in semi-supervised segmentation.

Findings

Improved segmentation accuracy on MRI datasets.

Effective semi-supervised learning with limited labeled data.

Addresses data augmentation challenges in segmentation.

Abstract

Recently proposed techniques for semi-supervised learning such as Temporal Ensembling and Mean Teacher have achieved state-of-the-art results in many important classification benchmarks. In this work, we expand the Mean Teacher approach to segmentation tasks and show that it can bring important improvements in a realistic small data regime using a publicly available multi-center dataset from the Magnetic Resonance Imaging (MRI) domain. We also devise a method to solve the problems that arise when using traditional data augmentation strategies for segmentation tasks on our new training scheme.