Carsten L

icon indicating a website link sten2lu.github.io

icon company PhD Student @Interactive Machine Learning Group; DKFZ / Helmholtz Imaging icon location Heidelberg icon location Machine Learning Research in Active Learning with past experiences in Uncertainty Prediction, Anomaly Detection and Generative Modeling.

Results 44 comments of Carsten L

For tooth segmentation from CBCT(labeling issue)

Hi @deepsea920415, nnU-Net can only be trained on datasets where an identical label is used for each pattern. E.g. - incisor tooth is always label 1 - molar tooth is...

The train_loss and val_loss synchronous rise.

Hi @CodeHarcourt, it seems to me as if your train loss is not steadily increasing but actually decreasing as the values are generally becoming more negative. Train loss epoch 0:...

RuntimeError: Some background worker is 6 feet under.

Hi @CYYJL and @avBuffer, The error message suggests reducing the number of workers as you are running out of RAM. You can do this by running with the flag np....

The mean validation dice is significantly lower than the pseudo dice.

nnU-Net is trained with patches of the images which are samples from the original images. These patches are also used to calculate the pseudo-dice metric during training. After training, nnU-Net...

When I train,it always stops at epoch 0

Hi @liyang19971022, I do not understand what is going on in your setting. Could you send me the nnU-Net commands that you are using for training alongside your error message?...

Improve segmentation in a specific region

HI @Laurabc, First, regarding the general approach: I encourage you to oversample specific regions during nnU-Net training to increase their weighting during training. For this to work, you would need...

Improve segmentation in a specific region

Hi @Laurabc, has your issue been resolved? Best regards, Carsten

I want to create a python class for inference

Hi @Body123 There is already a Python class for inference in nnU-Net. See here for an example: https://github.com/MIC-DKFZ/nnUNet/tree/master/nnunetv2/inference Best regards, Carsten

Did not find the expected number of training cases however I have 10 training cases (snapshots of error and json attached)

Hi @MaimoonaAkram. ## Issue Summary The script nnUNetv2_plan_and_preprocess is throwing an AssertionError because it did not find the expected number of training cases (10) in the dataset directory. Instead, it...

3D segmentation issue : IndexError: index 2 is out of bounds for axis 0 with size 2

Hi @kgh6784, Sorry for the late reply! The issue is that nnU-Net expects the data to be formatted without a channel dimension. Therefore: IMG: (depth, 256, 256, 1) would need...