Joel Akeret

Don't we have to train two models then? I wasn't aware that dropout is so time consuming. How much does it affect training/validation performance?

An 16% performance improvement is nice. However, i still don't fully understand how the training/validation procedure would look like. If a new model is created for validation, how would you...

I don't see how this should be implemented. The computation-graph would be different for the two networks, which makes it hard to transfer the weights from one to the other

it's probably a bit late, but the reason for this behaviour might be because the images are not preprocessed as the ones used for training (e.g. scaling to [0,1), removing...

Hmm odd that the loss is the same for all mini batches. Does it change after a few iterations? Do you get the same result in every time you run...

The loss is unnaturally high and should decrease every epoch. Given that the loss is always the same, independet of the net architecture I suspect that something might no be...

> 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...

What is the reason to change the spelling?

How are you reloading the model? Is the prediction working when using the [predict function](https://tf-unet.readthedocs.io/en/latest/tf_unet.html#tf_unet.unet.Unet.predict) of the model?

When creating a new instance of `Unet` the global graph is resettet using `tf.reset_default_graph()` to avoid problems with previous graphs. Not sure if this is causing the issue you're seeing