How to use different sized images to train a CNN

I need to train a CNN for image category classification of vehicle images, the images in data set that I have are of different sizes, and according to my knowledge we have to use a data set of same size for the image input layer, my questions are:

1. how can I use different sized image data set in CNN?
2. can this issue be resolved by image padding? and how to do it?
3. will padding reduce the accuracy of trained CNN?

thanks!

Answer 1

[Abolfazl Nejatian]

i've the same question !!!