search

rixez / Brats21_KAIST_MRI_Lab

Codes and pretrained weights for winning submission of 2021 Brain Tumor Segmentation (BraTS) Challenge
Apache License 2.0
169 stars 22 forks source link

[ENH] Paper Implementation using Tensorflow/Keras #1

Closed miladsade96 closed 2 years ago

miladsade96 commented 2 years ago

Hi @rixez

I would like to implement this paper using tensorflow and keras library. I have a problem with axial attention section and i can not understand it in details. Could you please give me some hints and/or resources to find out how to implement the axial attention?

Regards

rixez commented 2 years ago

Hi @EverLookNeverSee, Thank you for your interest in this repo! The axial attention in this package was adopted from Axial Attention. For the reference paper, you can check out the paper. The implementation of axial attention in this repo can be found here: https://github.com/rixez/Brats21_KAIST_MRI_Lab/blob/d05fba87d392f4ab9cc6c64c1658688f7d27a3d0/nnunet/network_architecture/generic_UNet.py#L377 And the trainer for the axial attention model can be found here: https://github.com/rixez/Brats21_KAIST_MRI_Lab/blob/d05fba87d392f4ab9cc6c64c1658688f7d27a3d0/nnunet/training/network_training/competitions_with_custom_Trainers/BraTS2020/nnUNetTrainerV2BraTSRegions_moreDA.py#L363 To train the model, you can replace the trainer in the readme with the appropriate trainer like this: nnUNet_train 3d_fullres nnUNetTrainerV2BraTSRegions_DA4_BN_BD_axialattention_unet <TASK_ID> <FOLD> --npz

I think the syntax would be similar in tensorflow. But note that we did not use the axial attention model in the final ensemble for the competition. Thank you and let me know if you have any more questions!

miladsade96 commented 2 years ago

@rixez Thanks for the explanation. If i have a question, i will ask you.