How to Dynamically Update ROI in DeepStream Without Restarting the Application?

Hi Community,

I am currently working on a DeepStream pipeline using DeepStream SDK 6.2 and need to dynamically update the Region of Interest (ROI) without restarting the application. Specifically, I want to:

  1. Adjust the ROI parameters (e.g., top, left, width, height) during runtime.
  2. Apply the updated ROI for inference using the nvinfer plugin.

I have seen discussions about defining multiple ROI filters in configuration files, but it is unclear how these can be modified dynamically while the pipeline is running.

Questions:

  • Is there a way to update ROI dynamically using DeepStream SDK 6.2?
  • Are there any API methods or plugins that can facilitate this process?
  • If not directly supported, are there recommended workarounds to achieve this?

Any guidance, examples, or links to relevant documentation would be greatly appreciated!

Thank you in advance for your help.

Hardware: A30
Driver: 535.183.01
Cuda: 12.2
DeepStream: deepstream:6.2

No. Currently not even with latest DeepStream SDK.

The Gst-nvdspreprocess (Alpha) — DeepStream 6.2 Release documentation is introduced to customize ROIs for nvinfer. It is open source, please customized your own preprocessing functions.

No workaround. DeepStream only support analytics ROI dynamic change. The inferencing ROI dynamic change should be implemented by yourself.

1 Like