had fun with mlmodel, drawable uiimageview and hiragana
This test is higly inspired by an article from Melody Yang
However the drawable implementation for writing the Hiragana was not the best Idea.
Using only Bezier path for your drawable area will result in severe lack of performance if user start drawing something else and will result, in the end, into a crash of your app. note quite what you'd like I guess.
Plus, to avoid caputring a UIView, to transform it in a UIImageView, I'm using a UIImageview from the begining.
DrawableCanvas can be used by whoever need a drawable suface, it just needs you to implement DrawabaleCanvasDelegate
Not quite what expected, model is probably not train enough or i'm doing mistakes with image tranformation to bytes. most of the time, regading less the hiragana you just draw け will be recognized.
readdme made using stackedit