About New OCR model’s improvement

[cyhasuka]

Concerned that yesterday huggingface had an update on the latest OCR model.

I performed a comparison test, including file slicing, GENERAL mode, and MANUAL mode embedding. But I am puzzled that I didn't find any change in the OCR results after updating the model, may I ask what are the improvements in the updated model?

Since OCR is critical for document recognition accuracy, are there any plans for further optimization? Thx.

[KevinHuSh]

Misoperation, ignore it please.
Yes, we're researching.

[yingfeng]

New document layout analysis model is introduced, see #4054

[cyhasuka]

That's Great!