search

MIC-DKFZ / nnDetection

nnDetection is a self-configuring framework for 3D (volumetric) medical object detection which can be applied to new data sets without manual intervention. It includes guides for 12 data sets that were used to develop and evaluate the performance of the proposed method.
Apache License 2.0
553 stars 97 forks source link

cross validation split #145

Closed EvelyneCalista closed 10 months ago

EvelyneCalista commented 1 year ago

hi @mibaumgartner ,

I got this warning when training the fold-0 which there is no positive sample, since my data is imbalance, i would like to to stratify split k-fold. however i cannot find the code line to split the cross validation. May I know how can I do stratify kfold for training?

_/opt/conda/lib/python3.8/site-packages/sklearn/metrics/_ranking.py:811: UndefinedMetricWarning: No positive samples in y_true, true positive value should be meaningless warnings.warn("No positive samples in ytrue, "

Thank you for your help.

mibaumgartner commented 1 year ago

hi @EvelyneCalista, you can always manually create a "splits_final.pkl" in the preprocessed directory of your data to define custom splits. However note: Since we are in object detection, the warning can have different meanings and does not automatically mean that there are not objects in your current validation. If the model is still at the beginning of the training phase it might not detect any object and thus leading to no positive samples on the object level evaluation.

EvelyneCalista commented 1 year ago

hi @mibaumgartner ,

thank you for your fast response and help. I understand now, I got wrong concept for positive & negative samples and it looks like i can ignore the warning since it is in the first epoch. sorry for the inconvenience, I am a beginner in object detection therefore I have lot of questions and confuses.

Thank you again

github-actions[bot] commented 10 months ago

This issue is stale because it has been open for 30 days with no activity.

github-actions[bot] commented 10 months ago

This issue was closed because it has been inactive for 14 days since being marked as stale.