Can this handle small images? #7
Description
Hello author, I would like to ask that Is this binarization method applicable to the image of the text line? For example: does a 40300 text line can work? I directly used your network weights to test such a small text line and found that the result was almost entirely white. Do you think it will work if I turn the training data into small text lines? I want to set the global image to 60500 and the local image to 60*60 width. Do you think this is ok ? Looking forward to your reply, thank you very much!