• Hardware Platform (Jetson / GPU)
• DeepStream Version
• JetPack Version (valid for Jetson only)
• TensorRT Version
nvdsmeta.h where it states "For objects found by a tracker and not inference component, confidence will be set to -0.1"
Would it be possible to change this such that the parameter contains the value from the last component to process confidence?.
If an IOU or KTL Tracker comes after the inference, to somehow carry the confidence value between frames that were inferred on? Perhaps this is a IOU and KTL update?
The challenge with the current solution is… if a client is making decisions in a Pad Probe callback based on confidence levels – i.e. shade the background of the object with the highest confidence – , then the client has to carry this information forward between inferred on frames.
It seems that it would be a lot easier if the Plugin did this instead
A feature request perhaps?