🔤 [Question] Does PaddleX 3.0 OCR Pipeline Support Arabic Language?

[moazizbera]

Hello PaddlePaddle Team and Community 👋,

I’m currently using PaddleX 3.0 and trying to build an OCR pipeline that supports both Arabic and English.

While English works as expected, Arabic text is not being detected or recognized properly. I suspect the default recognition model doesn't support Arabic.

Questions:
Does PaddleX 3.0 OCR pipeline natively support Arabic?

If not, can I integrate Arabic-trained models from PaddleOCR (like arabic_rec and ch_PP-OCRv4_det) with PaddleX?

Can anyone kindly share a working code example that uses PaddleX (or a hybrid approach with PaddleOCR) to extract Arabic text from images?

My goal is to process documents with both Arabic and English text using the PaddleX pipeline if possible. Any help, references, or shared samples would be highly appreciated.

Thank you for your time and efforts in building such great tools!

Best regards,
Mohamed Abdelrahman
maziz.abdelrahman@gmail.com