Some OCR mismatches

[madmalkav]

Some OCR mismatches

[madmalkav]

This kanji was added to the list of Jouyou Kanji on 2010. IMHO recognizing all the Jouyou kanjis is basic for an OCR, so I will open a ticket for PaddleOCR (I’ll maybe try to see if there is some other Jouyou kanji missing in their model if I can figure out how to script that 😅)

in any case, out of YomiNinja scope, so feel free to close the ticket

PS: for reference, PaddlePaddle/PaddleOCR#11249

[matt-m-o]

You're correct, it seems that 喉 wasn't included in the training data for the OCR model used for Japanese, the latest one provided by PaddlePaddle.
Unfortunately, this issue can't be quickly resolved, as it would require training a custom model. However, once we implement support for other OCR engines like Google Lens/Cloud Vision, this problem should become less prominent.