This repo is the official implementation for UCTNet.
Install packages
cd nnUNet
pip install -e .
cd UCTNet
pip install -e .
Download datasets BCV(Synapse), ACDC and ISIC2018. The dataset partitioning of BCV(Synapse) and ACDC follows TransUNet and the ISIC 2018 is divided according to dataset_ISIC2018.json.
UCTNet_BCV -dataset_path DATASET_PATH
UCTNet_ACDC -dataset_path DATASET_PATH
UCTNet_ISIC2018 -dataset_path DATASET_PATH
BCV
UCTNet_train -task 17 --fold 0 --custom_network UCTNet_3D -ei UCTNet --deterministic -pretrained_weights "/UCTNet/pretrained_weight/BCV/model_final_checkpoint.model"
ACDC
UCTNet_train -network 2d -task 27 --fold 0 --custom_network UCTNet_2D -ei UCTNet --deterministic -pretrained_weights "/UCTNet/pretrained_weight/ACDC/model_final_checkpoint.model"
ISIC2018
UCTNet_train -network 2d -task 100 --fold 0 --custom_network UCTNet_2D -ei UCTNet --deterministic
BCV
UCTNet_predict --task_name 17 --f 0 -ei UCTNet -chk model_final_checkpoint
ACDC
UCTNet_predict --task_name 27 -m 2d --f 0 -ei UCTNet -chk model_final_checkpoint
ISIC_2018
UCTNet_predict --task_name 100 -m 2d --f 0 -ei UCTNet -chk model_final_checkpoint
Part of codes are reused from the nnU-Net and PHTrans. Thanks to Wentao Liu and Fabian Isensee for the open source code.