Read Chinese Introduction：Here！

• 2021.7.7 The train and test codes are released.

• pytorch1.7
• torchio<=0.18.20
• python>=3.6

• You can modify hparam.py to determine whether 2D or 3D classification and whether multicategorization is possible.
• We provide algorithms for almost all 2D and 3D classification.
• This repository is compatible with almost all medical data formats(e.g. png, nii.gz, nii, mhd, nrrd, ...), by modifying fold_arch in hparam.py of the config.
• If you want to use a multi-category program, please modify the corresponding codes in data_function.py by yourself. I cannot identify your specific categories.

if your source dataset is :

categpry-0
├── source_1.png
├── source_2.png
├── source_3.png
└── ...

categpry-1
├── source_1.png
├── source_2.png
├── source_3.png
└── ...

then your should modify fold_arch as *.png, source_train_0_dir as categpry-0 and source_train_1_dir as categpry-1 in hparam.py

• without pretrained-model

set hparam.train_or_test to 'train'
python main.py

• with pretrained-model

set hparam.train_or_test to 'train'
set hparam.ckpt to True
python main.py

• testing

set hparam.train_or_test to 'test'
python main.py

• 2D

• alexnet
• densenet
• googlenet
• mobilenet
• nasnet
• resnet
• resnext
• vggnet

• 3D

• densenet3d
• resnet3d
• resnext3d

• dataset
• benchmark

This project is not perfect and there are still many problems. If you are using this project and would like to give the author some feedbacks, you can send Kangneng Zhou an email, his wechat number is: ellisgege666

This repository is an unoffical PyTorch implementation of Medical segmentation in 3D and 2D and highly based on pytorch-cifar100 and torchio.Thank you for the above repo. Thank you to Cheng Chen and Weili Jiang for all the help I received.