Datasets:
Tasks:
Image-to-Text
Modalities:
Image
Formats:
imagefolder
Languages:
Hebrew
Size:
10K - 100K
License:
Too much easy synthetic data
#2
by johnlockejrr - opened
Too much naive synthetic data. There are better synthetic data outthere, but only synthetic data is no real use for OCR. Good try anyway.
EDIT: tesseract and surya2 could work on perfect synth or good print. Try them on Hebrew manuscripts! ;)