I am working on ocr for Identity cards, my ocr( using YOLO) is quite good, but facing one challenge, when id’s are rotated (0-360), my model is not doing good. I want to train a model which can detect angle of rotation( same like google lens, which automatically detects the angle of rotation first then it performs the ocr.
Is there any model for detecting the angle of rotation ( normal methods like HoughLines) is not consistent with my requirements.
here is an example of few of the ids, I am working on
