I'm getting literally crazy trying to understand how U-NET works. Maybe it is very easy, but I'm stuck (and I have a terrible headache). So, I need your help.

I'm going to segment MRI to find white matter hyperintensities. I have a dataset with MRI brain images, and another dataset with the WMH. For each one of the brain images, I have one black image with white dots on it in the WMH dataset. These white dots represent where is a WMH on its corresponding brain image.

This is an image from the MRI brain images:

enter image description here

And this is the corresponding WMH image from the WMH dataset:

enter image description here

How can I use the other images in network validation?

I suppose there will be a loss function and this network is supervised learning.

  • $\begingroup$ first hit when googling "unet example" github.com/zhixuhao/unet will give you an idea how to use unet. you will need to be very familiar with your data, how to deal with the images and masks just to get it right. in my experience, unets work pretty well out of the box. you may run into questions like "how big of an image to feed in", "memory problems", etc. you'll need to take those one at a time. $\endgroup$ – user1269942 Dec 12 '19 at 20:04
  • $\begingroup$ Have you solved this problem? It's not clear exactly what you're asking. Are you asking how to augment your training data? $\endgroup$ – nbro Jun 12 '20 at 23:57
  • $\begingroup$ @nbro I think I was asking about validation data parameter used in model.fit method. Six months later, I have understood what I have to do (I think). Maybe the question is about how to use validation data and how to use the method predict to validate training. $\endgroup$ – VansFannel Jun 13 '20 at 8:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.