NvDsObjectMeta confidence value

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) AGX Xavier
• DeepStream Version 5.0
• JetPack Version (valid for Jetson only) 4.4

I am running deepstream_test1 with the default dstest1_gie_config.txt configure file. It runs well. But when I try to print out (NvDsObjectMeta) obj_meta->confidence and I see all the detected objects with the same confidence of -0.1. Is this normal/correct? I am expecting positive confidence with different value for each detected object of floating point number between 0 to 1. Please advise how to read out the object confidence level and what value to expect. Thanks a lot.

3

Please chech the comments in NVIDIA DeepStream SDK API Reference: Main Page

gfloat confidence
Holds a confidence value for the object, set by the inference component.

confidence will be set to -0.1, if “Group Rectangles” mode of clustering is chosen since the algorithm does not preserve confidence values. Also, for objects found by tracker and not inference component, confidence will be set to -0.1
4

Thank you for the information. I found the following description in dstest1_pgie_config.tx and no further explanation what other options are available:

cluster-mode(Default=Group Rectangles)

I wonder what are the other cluster-mode available to expose the object confidence. Could you please point me to the documentation for the nvinfer that provide description for other cluster-mode settings? Thank you very much for your help.

5

Found it: here

I set the following cluster-mode=2 in my dstest1_pgie_config.txt file:

cluster-mode=2 #0: OpenCV groupRectanges; 1: DBSCAN; 2: Non Maximum Suppression; 3: DBSCAN + NMS Hybrid; 4: No clustering

Then the obj_meta->confidence shows up as expected!

6

Great work.