Obj.confidence is negative?

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Hi All,

I am running Deepstream 5.0 on both my Jetson Nano & Jetson NX using python examples.
I found in the sample programs, the “obj_meta.confidence” is negative value such as -0.100000000149 etc.

Based on my understanding, the confidence value should be a value that betweeen [0, 1].

Why is the confidence negative?
I want to show the confidence onto the frame, how can I get the real confidence value that between 0, 1?

Thanks

1 Like
3

Hi,

The negative confidence is from the group rectangles type of clustering method.
This method does not output the confidence values for the clustered boxes and hence all objects will have -0.1 conf at the pgie.

Please check this topic for the detail:

Thanks.

1 Like
4

I saw there will be a patch or update about this issue, which will solve my problem?

Am I correct?

5

Hi,

The confidence -0.1 is set by clustering approach since the confidence value is not available.
The patch in next release will separate the confidence value for pgie and tracker to avoid confusion.

Thanks.

6

When will the next patch be available?

7

Hi,

Sorry that we cannot disclosure our future plan here.
Please pay attention to our announcement for the update.

Thanks.

9

Hi , I have the same problem , when I run the application in my xavier AGX every time the meta.confidence have the same value and it’s négatif , -0.10 .
When I see visualisation of inference in Jupiter , I don’t have this problem , all it’s OK .
Do you know how I can fix it or if we need to wait update from you ?
Thank you

10

Hi lyron,

Please help to open a new topic for your issue. Thanks