No description on Normalization

Hi you have used the following line of code in object_detector_detection_api_lite.py,

frame = (2.0 / 255.0) * frame - 1.0

Why do we do that? Also if we further quantize the model, would this line change?
