search

CSCYQJ / MICCAI23-ProtoContra-SFDA

This is the official code of MICCAI23 paper "Source-Free Domain Adaptation for Medical Image Segmentation via Prototype-Anchored Feature Alignment and Contrastive Learning"
24 stars 6 forks source link

About the data preprocessing #12

Open derekshiii opened 8 months ago

derekshiii commented 8 months ago

The work of the paper is very impressive I would like to ask about the preprocessing of your dataset. I have seen that your dataset is not the original BTCV and CHAOS dataset in other issues, so I would like to ask what is the preprocessing of these data

BarY7 commented 4 months ago

Hi @derekshiii , did you manage to recreate the preprocessing? It seems the preprocessing used is similar to SIFA:

https://github.com/cchen-cc/SIFA?tab=readme-ov-file

But both in SIFA and in my implementation I don't seem to reach the same results as the paper reports

Milk-White commented 3 months ago

Hello@BarY7 , I downloaded the dataset used in SIFA and found that only the test set ends in. npz, while the training set is in other formats. May I know how your training set was processed, or do you have a link to a processed MR training set? Thank you!

BarY7 commented 3 months ago

@Milk-White You can try looking in the preprocess notebook in the commit history. I pretty much gave up because even with the notebook its hard to reproduce the author's results, and they are not available here sadly

Milk-White commented 2 months ago

@BarY7 Thank you for your answer, thank you