Cost_function_config for Detectnet_v2

How does the following parameters in cost_function_config affect the model?

  1. coverage_foreground_weight
  2. initial weight and final weight for both coverage and bbox loss.
  3. what should be the ideal values for these?
  4. On what basis do we tweak these values?
    Currently it looks like trial and error with no logic to base these values on.

Refer to

and
https://docs.nvidia.com/metropolis/TLT/tlt-user-guide/text/faqs.html#training