Ask HN: What open source OCR models are available?

Posted by zerojames on Saturday May 13, 2023 @09:10AM
from the ask-hn dept.

I have tried Tesseract for OCR but I wondered if there are any more accurate tools available. I have been experimenting with some Visual Question Answering (VQA) models such as BLIPv2 but they generally don't return full OCR results, rather a limited set of the words in an image.

( 3 comments | 3 points | on HN )

Re:Ask HN: What open source OCR models are available? by itg on Saturday May 13, 2023 @09:18AM (#35928082)

I only know of Tesseract and PaddleOCR if you want to run them locally.

[ Parent | on HN ]

Re:Ask HN: What open source OCR models are available? by rainytuesday on Saturday May 13, 2023 @10:06AM (#35928518)

https://huggingface.co/docs/transformers/model_doc/trocr

[ Parent | on HN ]

Re:Ask HN: What open source OCR models are available? by IndigoIncognito on Saturday May 13, 2023 @07:16PM (#35933740)

The ones i've tried are pretty terrible compared to google's OCR which is really good but closed sauce

[ Parent | on HN ]