No recognition of slavic characters. English characters recognized are separate singular characters, not a block of text.

[MortalWombat-repo]

I am using paddleOCR as a fastapi server on huggingface spaces free tier, without gpu, only 2 cpu cores.
I don't know whether that is a limitation?

This is the repo
Link

It can be accessed with
curl -X POST -F "[email protected]" https://icosar-ocr-api-paddleocr.hf.space/ocr
as it is open.

I am using this image.

And I get this output:
{"text":["n","a","o","t","o","e","e","e","e","e","e","e","e"],"message":"Text detected"}

I would be most appreciative of any guidance.

Tessaract 5 is much more accurate, and I suspect an error on my part.

[liuhongen1234567]

Hello, I am getting satisfactory results using PP-OCRv5 for OCR recognition, and PaddleOCR runs quite fast on the CPU. Perhaps you could install PaddleOCR 3.0 locally?

Alternatively, you can use the official PaddleOCR online experience, such as the AI Studio PP-OCRv5.
https://aistudio.baidu.com/community/app/91660/webUI

[zhangyubo0722]

Can you share how you are using it?