search

sergivalverde / nicMSlesions

Easy multiple sclerosis white matter lesion segmentation using convolutional deep neural networks.
GNU General Public License v3.0
46 stars 21 forks source link

Overfitting during retraining #21

Closed letilessa closed 1 year ago

letilessa commented 3 years ago

Dear Sergi, I full trained a model using ISBI and MICCAI dataset and got very good results. But now, I am trying to use this model for retraining the last 3 FC layers as the domain adaptation with the target images, however I am having trouble with overfitting. What would be the optimal number of epochs and patience?

Kind regards

ghofranGh2022 commented 2 years ago

Hi @letilessa has the nicMSlesions approach been carried out with you ??, can you help me please, I have a lot of problems doing the training and the test like this: Thanks. Capture22

sergivalverde commented 2 years ago

Hi @ghofranGh2022

Are you pointing the configuration options to the folder where data is?

This looks suspicious:

PRE: .ipynb_checkpoints identifying input modalities

sergivalverde commented 2 years ago

@letilessa Can you share your experience? Did it work for you finally?

ghofranGh2022 commented 2 years ago

Mr.Sergi @sergivalverde , I don't know how to do that!!, I just created a folder that contains flair, t1, lesion data, I converted them to nifti format, then I copied the path in the configuration.cfg file and the default.cfg file!