Training classification model for face detection

I am trying to train my own classification model for face detection using trainingImageLabeler. Is that possible to use images with different formats to train the classifier or should be the same? and what is the best size for the training images?

Thank you.