Multi-objects recognition with single object tracking on Deepstream

Accelerated Computing Intelligent Video Analytics DeepStream SDK
, ,
1

Hello,

I have built a multi-object recognition model using Deepsteam python bindings and YOLO and I currently have a tracker attached to it (modified from deepstream_test2 python example), and it’s is working fine. However, I want to attach the tracker to one specific class only and not all the detected classes (e.g. Person only). Is there a way to achieve this ?

• Hardware Platform (Jetson / GPU) Jetson Xavier AGX
• DeepStream Version 5.1
• JetPack Version (valid for Jetson only) 4.5

Thanks

3

Hi,

You can filter the output bbox from detector directly.

https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_plugin_gst-nvinfer.html#gst-nvinfer-file-configuration-specifications

Property Meaning Type and Range Example Network Types
filter-out-class-ids Filter out detected objects belonging to specified class-ids Semicolon delimited integer array filter-out-class-ids=1;2 Detector

Thanks.

4

Thanks for you replay. I essentially do not want to filter from the detector, but from the tracker instead. Say I am detecting 4 objects and would want to track only one of them.

1 Like