Issue with poor detection of TLT-converted network in DeepStream

I have a Faster R-CNN network that has good detection outside of DeepStream but when I try to TLT convert it to an engine and use it in DeepStream, it is not able to detect the same objects well.

Here is my DeepStream config file:
kv_test_config.txt (2.4 KB)

Thank you for any assistance you can provide.

How did you convert it to an engine? Can you share more information?

There is no update from you for a period, assuming this is not an issue any more.
Hence we are closing this topic. If need further support, please open a new one.

Do you mean the model can get good detection with tlt faster inference ?