Add scripts for generating train/test datasets, determining normalization mean and std, and min/max resolutions

[jakobj]

As the title suggests.

These scripts assume that the original dataset is located in <root>/data/dataset.