How to prepare my own training data

Hi, I would like use this to detect objects for autonomous cars. I have collected my own images. I can also label my images by drawing the boundingbox of cars, person, poles, signs, bikes, etc. But in what format should I write this label data?