jakeret/tf_unet

question

pierpy opened this issue · 3 comments

pierpy commented

jDear Doctor Akeret,
I found on github your interesting u-net python tool. I'm trying to use it in a project.
I can train the network on my data but I have two questions:

  1. If i try to increase the number of layers I get an error;
  2. Why the origninal image is cropped? Is it possible to train on the whole immage?
  3. On the contrary, is it possible to train considering patches of the original image?

Thanks in advance

jakeret commented

If i try to increase the number of layers I get an error;

Without description we can't help you

Why the origninal image is cropped? Is it possible to train on the whole immage?

This is due to the network architecture using "valid" padding in the convolutional layers - you can read more on this in the original Ronneberger et al. paper

On the contrary, is it possible to train considering patches of the original image?

Yes, but you have to implement this yourself

pierpy commented
daboe01 commented

@pierpy split the images to your needs before placing them into the training folder