search

vios-s / vMFNet

vMFNet: Compositionality Meets Domain-generalised Segmentation
MIT License
16 stars 3 forks source link

Compared with the indicators #2

Open wangmingaaaaa opened 1 year ago

wangmingaaaaa commented 1 year ago

Hi, I would like to ask why your paper is not related to the Enhancing Pseudo Label Quality for Semi-Supervised Domain-Generalized in 2022 Compared with the indicators of Medical Image Segmentation, is it because the backbone is different?

xxxliu95 commented 1 year ago

Hi, yes, the paper you mentioned used a pre-trained ResNet as the backbone.

See in the related work: Following \cite{liu2021semi}, Yao et al.\ \cite{yao2022enhancing} adopted a pre-trained ResNet \cite{he2016deep} as a backbone feature extractor and augmented the source data by mixing MRI images in the Fourier domain and employed pseudo-labelling to leverage the unlabelled data.

Different backbones affect the DG performance.