Deepstream Python drop frames dynamically to keep inference real time

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

Is it possible to set config parameters in the Python Deepstream reference apps to keep the inference real time (i.e. keep processing real time frames) by dynamically dropping frames?

3

Hi,

YES. There are some configure you can try.

1. Use fp16 model.
Ex. dstest1_pgie_config.txt

[property]
...
## 0=FP32, 1=INT8, 2=FP16 mode
network-mode=2
...

2. Run inference periodically and generate the intermediate output via tracker
https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_plugin_gst-nvinfer.html#gst-nvinfer-file-configuration-specifications

[property]
...
interval=0
...

Thanks.