search

ellisdg / 3DUnetCNN

Pytorch 3D U-Net Convolution Neural Network (CNN) designed for medical image segmentation
MIT License
1.9k stars 653 forks source link

Valued Error on BraTS20 Training #307

Closed tarekegn82 closed 1 year ago

tarekegn82 commented 1 year ago

Dear Ellis,

Thank you for your contribution. I have been trying to run a training experiment on the BraTS20 dataset which I got from Kaggle and I couldn't make any progress due to "ValueError: num_samples should be a positive integer value, but got num_samples=0". I put the dataset in the folder. /brats2020 and didn't modify the scripts further. Please kindly share your suggestion.

Best

ellisdg commented 1 year ago

That error means that the image filenames aren't being found. I'll try to get a configuration file working tomorrow.

tarekegn82 commented 1 year ago

Sure, thank you very much. Please give me updates here when you're done.

ellisdg commented 1 year ago

Check out the new brats2020 tutorial/example. I was able to get it running this morning using the dataset from Kaggle. I'm going to close this issue, but feel free to open it if it doesn't work for you.

tarekegn82 commented 1 year ago

It is fixed as you can see from the attachment. Thank you for your time.

On Thu, Aug 3, 2023 at 12:18 AM David G Ellis @.***> wrote:

Check out the new brats2020 tutorial/example. I was able to get it running this morning using the dataset from Kaggle. I'm going to close this issue, but feel free to open it if it doesn't work for you.

— Reply to this email directly, view it on GitHub https://github.com/ellisdg/3DUnetCNN/issues/307#issuecomment-1662519222, or unsubscribe https://github.com/notifications/unsubscribe-auth/APZ6CFY4S5EZNIDXVZ4DMK3XTJ4U3ANCNFSM6AAAAAA26VVHHE . You are receiving this because you authored the thread.Message ID: @.***>

ellisdg commented 1 year ago

I realized that I had the labels wrong in the configuration file and setup. The labels should be [2, 1, 4]. I got confused as I was going back and forth between brats 2023 and brats 2020. This should be fixed now in the latest PR #310.

tarekegn82 commented 1 year ago

Noted with thanks. I also have brats2021 Train and Val dataset, so I’ll check the model with it as well, I mean with the earlier version. But brats2021 is much larger dataset in terms of train and val cases. So, not sure this would raise any issues.

On Thu, Aug 3, 2023 at 9:47 AM David G Ellis @.***> wrote:

I realized that I had the labels wrong in the configuration file and setup. The labels should be [2, 1, 4]. I got confused as I was going back and forth between brats 2023 and brats 2020. This should be fixed now in the latest PR #310 https://github.com/ellisdg/3DUnetCNN/pull/310.

— Reply to this email directly, view it on GitHub https://github.com/ellisdg/3DUnetCNN/issues/307#issuecomment-1663178039, or unsubscribe https://github.com/notifications/unsubscribe-auth/APZ6CF7II5MSG2B6SJV3243XTL7JNANCNFSM6AAAAAA26VVHHE . You are receiving this because you authored the thread.Message ID: @.***>

--

TAREKEGN