Hi,
how feasible would be to use a custom .tflite trained model? Seems like MLKit allows it. I'm using this package to scan document's MRZ lines. Already with good results, but a MRZ trained model would rock.
On another hand, on MlkitOcrModule.kt I see TextRecognizer is using latin lang by default. Would be DOCUMENT (or maybe CREDIT_CARD] more suitable for MRZ -or, at least, a little step forward-? In that case, would be lang option easier to add to detectFrom*() , maybe? (Dunno on the ios side, though).
Thanks!