small-models-for-glam/kraken-ppocrv6-medium

Image-to-Text • Updated 6 days ago • 13

Note Hub mirror of Benjamin Kiessling’s multilingual Kraken recogniser for print and handwriting; needs Kraken 7.1+ and line segmentation. Usage: https://zenodo.org/records/21788410

PaddlePaddle/PP-OCRv6_medium_rec

Image-to-Text • Updated Jun 12 • 55.6k • 29

Note 19M-parameter multilingual text recogniser; supply crops or use PaddleOCR detection for whole pages. Usage: https://github.com/PaddlePaddle/PaddleOCR

kha-white/manga-ocr-base

Image-to-Text • Updated Jun 22, 2022 • 945k • 180

Note Japanese manga text, including vertical writing and furigana; supply one text region, since blank images can produce invented text. Usage: https://github.com/kha-white/manga-ocr

JustANormalTinkerer/hayai-ocr-v2

Image-to-Text • 0.2B • Updated 3 days ago • 3.45k • 7

Note Small CJK/English recogniser for cropped, stylised and vertical text; the v2 repository hosts v2.1 and requires custom model code.

microsoft/trocr-base-printed

Image-to-Text • 0.3B • Updated May 27, 2024 • 214k • 218

Note Printed-English recognition with a short Transformers example; requires line crops and does not reconstruct page layout.

nvidia/nemotron-ocr-v2

Image-to-Text • Updated May 22 • 1.37k • 255

Note Detection, recognition and layout for documents and natural-scene images; use the full pipeline under the NVIDIA Open Model License.