I have a trained a model, and it is labeling the correct locations in the images. My issue is when I want to crop the labeled part so I can do OCR. I found the blog post about cropping and followed the steps, but the issue is the cropped part is not in the right location and it is very small. I am not sure why would that happen.