Tensorflow should I crop objects from the images for better accuarcy?

Question

I am new to tensorflow and I could not find an answer for my question.

I am trying to make a simple program what recognises the type of van from the picture. I downloaded