How to track ONLY Person using deepstream_sdk in Jetson AGX Xavier ?

Hi,
I am trying to detect objects using deepstream_sdk parallel in 6 cameras.
I am able to do that but now I want to detect ONLY Person no other objects like car, bicycle etc.

Is there any option to track Person only or disable other objects in deepstream-app?

Please help if any one have tried the same case with deepstream sdk.

Thanks

Hi,

Our primary network is trained to detected four class.
You will need to retrain the model for only one specific class output.

A possible workaround is to update the threshold value of unwanted to the maximal.
You can use per class configuration instead of class-attrs-all.
And set threshold to 1.0 to avoid any bbox generate.

[class-attrs-0]
threshold=1.0
group-threshold=1
## Set eps=0.7 and minBoxes for enable-dbscan=1
eps=0.2
#minBoxes=3
roi-top-offset=0
roi-bottom-offset=0
detected-min-w=0
detected-min-h=0
detected-max-w=0
detected-max-h=0

[class-attrs-1]
...

Thanks.

Hi AastaLLL,
Thanks for the valuable help.
Now problem solved.
One more doubt Can I get 60fps streaming in deep streamer if I used 6 YUV cameras parallelly?

I got 60fps in deep streamer in case of 4 cameras but not able to get 60fps in case of 6 cameras.
anyways to tune this issue other than nvpmodel.
Thanks

Hi AastaLLL,

I would like to know how I can specify the nvtracker to only perform tracking on one or two classes,

I am using the deepstream python bindings, is there a property that can be set on the tracker?