I want to extract the text from an id card. The text can be even rotated also. I want to annotate the image with help of rotating bounding boxes.
My doubt is should the images have to be same size before giving as input to yolo?
Inputs in the same shape make it often easier to feed them as batches to the model.
Otherwise you would have to resize or pad them in order to create batches.
If you have the ability to resize them now to a static shape, I would try it in case you don’t have strong arguments against it.