Model training overhaul

The model training feature is currently somewhat limited; once a model is trained, there isn't a way to continue training the model with additional images or many finetuning options.
I would like to be able to provide manual feedback/intervention after every epoch instead of setting the options at the start and hoping for the best.