search

clovaai / deep-text-recognition-benchmark

Text recognition (optical character recognition) with deep learning methods, ICCV 2019
Apache License 2.0
3.76k stars 1.11k forks source link

when ocr recognizes the low resolution images , should ocr model train and test on either the gray image or the rgb image ? #169

Open anh1nt opened 4 years ago

anh1nt commented 4 years ago

In general, using the gray image for ocr is good. However, when ocr recognizes the text images, which are low-resolution images and unclear text texture. should ocr model keeping those images as the color image (rgb) learning or converting them into the gray image. Do you think the color image may provide much more information in this case of low-resolution image recognition?

Thank you very much.

ku21fan commented 4 years ago

Hello,

Yes, I think that the color image could provide much more information. But in our experiment, the accuracy difference was trivial. In other words, empirically there was no big gap.

But in a particular situation, such as low-resolution images you mentioned, the color images could be better.

I have an interest in the results :)

Best

Charlyo commented 2 years ago

@anh1nt @ku21fan Has anyone trained the model with rgb in the branch recognition?

Charlyo commented 2 years ago

@ku21fan Could you share the color model? thkns!