Train and tune ESRGAN with Antarctic Snow Accumulation input

[weiji14]

Extends #146 to actually use the Antarctic Snow Accumulation dataset by Arthern et al. 2006 in our adapted Enhanced Super Resolution Generative Adversarial Network model!

TODO:

• [x] Establish a 'pre-' baseline without the Snow Accumulation input (7dec5b5cdfc5934454507b7a91d5bdab00001031)
• [x] Extend the DeepBedMap ESRGAN model to take in Snow Accumulation as input W3 (054e2953210f24844654ce78920d4afd834c5174)
• [x] Retune hyperparameters, set new hyperparameter defaults, update figures (83e956db3c36d9fd40f41e7d905889fa1fe08bfa, 75266fc181d8ed4961cb44cd17080b6a80a042ee, e8ae2740061424fd78e34c5bdacf9c6e852b6168)

[review-notebook-app[bot]]

Check out this pull request on ReviewNB: https://app.reviewnb.com/weiji14/deepbedmap/pull/151

Visit www.reviewnb.com to know how we simplify your Jupyter Notebook workflows.