HieuPhan33 / MaskGAN

MaskGAN for unpaired MR-to-CT synthesis. Official implementation of MICCAI 2023
27 stars 6 forks source link

About data #4

Closed HilbertMaximov closed 3 months ago

HilbertMaximov commented 9 months ago

Dear author, hello. I noticed that your data seems to be sliced from 3D nii and saved in image format. My datasets (CT and MR) are both in PNG format. Can you tell me the specific processing method? Also, could you describe the folder arrangement of the processed dataset? For example, for specific methods like datasets/trainA or trainB like Cyclegan, could you please take a screenshot if it's convenient? This can keep your private data confidential. Thank you very much!

HieuPhan33 commented 9 months ago

Hi sorry for the late reply,

We only need to apply preprocessing to convert 3D to PNG. If you already have PNG. Please go ahead to train the model.

The exampled data folder structure is as below:

train_A | ------- img1.png | ------- img2.png train_B | ------- img1.png | ------- img2.png val_A | ------- img3.png | ------- img4.png val_B | ------- img3.png | ------- img4.png test_A | ------- img5.png | ------- img6.png test_B | ------- img5.png | ------- img6.png