Tracker Confidence vs Inference Confidence

Deepstream Version 6.0

This question is about tracker and inference confidences on Deepstream 6.0.

I understand that inference confidence is available when using cluster-mode=1 or 2, (maybe 3?), and that tracker confidence is available when using the visual tracker (NvDCF).

How does tracker confidence relate to the inference confidence ? Is it confidence that the object has the correct id, or confidence that the object has been correctly classified, or both ?

In the case of wishing to extablish a filter to remove false positives caused by video noise, which confidence value would be most appropriate ?

The inference confidence comes from the model analysis and the tracker confidence comes from the tracking algorithm.

The inference confidence.

Thanks yuwiew for the prompt response.

Could you clarify what you meant by the answer you gave to the quoted text in your response ? Did you mean to answer “the inference confidence” to the question about video noise ?

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

Yes, you can try to use this to see if it works. Thanks

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.