Detectnet - deepstream params

I noticed that even slight changes in these parameters effect the perfromance of the model greatly. I am using a detectnet model trained on tlt.

The parameters are named below:


Hi ishan,

It seems you did not describe your issues clearly, please re-update your issue again, then we can know how to help you. Thanks

In the config file when I change pre-cluster-threshold by even .01 - it dramatically effects the accuracy of the model as well as the eps.

My issue is that I would love to get more information on these parameters so I can understand them more deeply and change them accordingly in the future.

The threshold applied to the final aggregate confidence values to render the bboxes. User can be tuned to any float value.

Hi @ishan
Looks you asked the same question in Pre cluster threshold

Does your quetion get addressed?