Follow UNet paper for training specifics

[thomashopkins32]

For the first couple of releases, I trained the model using things from my personal experience and intuition in training deep neural networks. I should look to see what worked for the authors and try to emulate that for this dataset.

This means that I should:

• Use SGD with momentum
• Use xavier normal weight initialization (check the math for this)
• Use data augmentation similar to their choices
• Use dropout regularization near the final layers of the network

Let's see if that will improve our performances.

[thomashopkins32]

Next step is to visually confirm that the elastic transforms work.

Then we need to implement the pixel weighting scheme based on class frequencies.