A Pseudo-labeling Approach to Semi-supervised Organ Segmentation

Jianwei Gao, Juan Xu, Honggao Fei

Keywords: Abdominal organ segmentation, Semi-supervised, Pseudo-labeling

Abstract: In this paper, we adopt a "pseudo-labeling" approach to semi-supervised learning based on 50 labeled data and 2000 unlabeled data. This approach yields a model with 0.7496 mean DSC on the validation set, outperforming the 0.6903 mean DSC of the model with only 50 labeled data.