I was wondering has my dataset is huge, if I crop the images to annotated parts, which would result in the 1 whole image = represent 1 object. will this affect model accuracy and overall performance?
Additional context
Has my dataset is large and I have limited training resources, I wanted to try different ways to reduce the training time without affecting the accuracy. So I had a plan to instead of providing a complete image that consists of a small part of the object to be detected why not just crop the image and give only the object part of the image.
❔Question
I was wondering has my dataset is huge, if I crop the images to annotated parts, which would result in the 1 whole image = represent 1 object. will this affect model accuracy and overall performance?
Additional context
Has my dataset is large and I have limited training resources, I wanted to try different ways to reduce the training time without affecting the accuracy. So I had a plan to instead of providing a complete image that consists of a small part of the object to be detected why not just crop the image and give only the object part of the image.