How to use custom trained Detectnetv2 resnet18 model in Gst-nvdsanalytics

I have trained a custom model using TLT2.0. The current Gst-nvdsanalytics config file has entries for model-file and proto-file. I removed these entries and put a model-engine-file=.

What other changes needs to be done. Do you have a sample please?

Can we also take output into an RTSP sink? If yes a sample would be helpful.

Hi @Mike
Yes, that’s ok, no other change needed for inference.

For the RTSP, you could refer to below question in FAQ - https://docs.nvidia.com/metropolis/deepstream/4.0/faq/index.html
How can I specify RTSP streaming of DeepStream output?

if you are using DS python , you can refer to the “deepstream-test1-rtsp-out” sample