Doesn't recognize at all

[websorokinweb]

from manga_ocr import MangaOcr

mocr = MangaOcr()
text = mocr('Shingeki No Kyojin - Raw/_Chapter 6/003.jpg')
print(text)

Got そして、日本では

I can't get what's wrong it's like common image + others are return something wrong too
Python 3.9.11
Windows 10

[websorokinweb]

I got. This package doesn't work for whole page. Only for bubbles

[HighLiuk]

@websorokinweb got the same problem here. I was so excited at first...

[HighLiuk]

@websorokinweb if you look at Mokuro, looks like it combines Comic Text Detector which is great to detect Japanese text and group them to bubbles, then for each bubble you use MangaOcr and voilà (at least in theory)

[HighLiuk]

@websorokinweb this should be fine:

from mokuro.manga_page_ocr import MangaPageOcr
from mokuro.utils import dump_json

if __name__ == '__main__':
    mpocr = MangaPageOcr()
    result = mpocr('path/to/image.jpg')
    dump_json(result, 'path/to/result.json')