Refactor; move consts to consts.py

gitignore the model

Adding visualize.ipynb, data_transform>RandomCrop

It does help if we save the model :-)

U gotta normalize. (+ cleanup)

Train a Font Identifier using ResNet18

Train and test splitter. Clean up comments and code a bit (but not too much).

Font allowlist, deal with .ttc files, let's just generate 10 per font

gen_sampla_data script to generate sample text images

First commit, just a README