The OCR bug returns a continuous string of zeros
when testing a handwritten font image in the space, the recognition works normally. However, after I deployed it locally according to your instructions, the recognition was abnormal, returning:
0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000\n
Interesting, I have no problems testing on handwritten notes in English locally (in the space I got very strange results though). Is your text in English? What's the size of the image? Can you share the image?
i have similiar issue where the model give output !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! in some images(mostly normal but somehow its bugged)