How to use the trained model to infer the raw nii.gz data?

[TongYoungG]

Dear Authors: Your job is excellent! Can you tell me how to use the trained model to infer the raw nii.gz data. The evaluation code seems to process the after-cropping data.

[Amshaker]

Hi @TongYoungG ,

Thank you so much for your interest in our work!

Yes, the evaluation code expects pre-processed data.

To infer raw nii.gz, you have to apply the same pre-processing and cropping we have here.

It will be better and easier if you followed the full steps of nnFormer for Setting up the datasets. Basically, you have to use convert_decathlon_task and plan_and_preprocess scripts to put the data into the correct format.

I hope it is clear now.

Best regards, Abdelrahman.

[TongYoungG]

Hi @TongYoungG ,

Thank you so much for your interest in our work!

Yes, the evaluation code expects pre-processed data.

To infer raw nii.gz, you have to apply the same pre-processing and cropping we have here.

It will be better and easier if you followed the full steps of nnFormer for Setting up the datasets. Basically, you have to use convert_decathlon_task and plan_and_preprocess scripts to put the data into the correct format.

I hope it is clear now.

Best regards, Abdelrahman.

Thank you very much!

[TongYoungG]

Dear Author: Sorry to disturb you. I meet two problems:

• [ ] When I train my personal dataset following the synapse network, the training is slow in the loading data stage data_dict = next(data_generator).
• [ ] when batchsize set 4, it gives a runtime error:

raise RuntimeError("MultiThreadedAugmenter.abort_event was set, something went wrong. Maybe one of " RuntimeError:

But batchsize=1 workes.

can you give me a hand?