Missing text with OCR although detected

[EHadoux]

Hey, I love the model, it's game changer.

I'm trying to extract this document. All is well except on page 6. As you can see on the screens (surya_gui, but it's the same in direct Python), the bounded boxes are good with "Run Text Detection", but the right-hand side of the first line is missing with "Run OCR".

The code I'm using which returns the same thing but directly in Python:

recognition_predictor(images, [["en"]] * len(images), detection_predictor, highres_images=images_high)

Any idea? Thanks!

12102682_2022-07-06.pdf

[kaiwang13]

Any solution for this case? I have met the same problem.