Tensorflow object_detection API fine tune ckpt classification

Question

I am just wondering what is the difference between the "classification" and "detection" fine tuning types for the available checkpoints in the object det