Fabian Isensee

icon location interested in deep learning with focus on medical image segmentation

Results 453 comments of Fabian Isensee

batch_SIZE

batch size is always 1 in inference and due to the sliding window implementation it's not simple to change that!

batch_SIZE

You cannot modify it while the training runs. Please follow this guide https://github.com/MIC-DKFZ/nnUNet/blob/master/documentation/tutorials/edit_plans_files.md

some problem about dataset_conversion for the Task114_heart_MNMs

We do not support windows - sorry :-/ @PeterMFull can you please change the conversion script so that it no longer relies on flsplit? I have a split_4d function somewhere...

Problem with predict

does this folder exit? If not we need to figure out why. What training command did you use?

Problem with predict

it should be `-tr nnUNetTrainerV2`. nnUNetPlansv2.1 is the plans identifier and doesn't have to be specified because it's the default anyways

Problem with predict

`nnUNet_predict -i /homedtic/czs/neuroendocrinetumors/nnUNet_raw_data_base/nnUNet_raw_data/Task001_tumors/imagesTs/ -o /homedtic/czs/neuroendocrinetumors/RESULTS_FOLDER/ -t 1 -m 3d_fullres -tr nnUNetTrainerV2` can you please try this?

Problem with predict

It tells you what folder it expects. Have you run the corresponding training? The command should be `nnUNet_train 3d_fullres nnUNetTrainerV2 1 0` for fold 0

Package version conflict

You should create a clean environment in that case. imageio is not part of nnU-Net so it's not a mistake on our end

Traning hangs every 100s of epochs

Hi, hard to say because that doesn't happen for me. What configuration are you training? And have you looked at your RAM? If the RAM gets full this can happen...

Bianry Segmentation for 3d ultrasound images doesn't work well

Hi there, this is really really hard to say without looking at the data. the most likely problem is that something went wrong during the dataset conversion. Can you please...