Why we need image pre and post processing when deploying a tlt model by using TensorRT?

[Morganh]

Refer to postprocessing in https://github.com/NVIDIA-AI-IOT/deepstream_tlt_apps/blob/master/post_processor/nvdsinfer_custombboxparser_tlt.cpp

Refer to preprocessing in https://github.com/NVIDIA-AI-IOT/deepstream_tlt_apps/blob/master/configs/ssd_tlt/pgie_ssd_tlt_config.txt
TLT the pre-processing of SSD image is like below:

• assume RGB input values in range from 0.0 to 255.0 as float
• change from RGB to BGR
• then subtract channels of input values by 103.939;116.779;123.68 separately for BGR channels.