Rotate input video in DeepStream python app

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) Jetson Nano
• DeepStream Version 6
• JetPack Version (valid for Jetson only) 4.6
• TensorRT Version 8.0.1.6

• Issue Type( questions, new requirements, bugs)
I have 4 rtsp sources and I want to rotate 1 source before performing inference. I have tried nvvideoconvert but it applies on all 4 sources instead of one.
Any solution please.

Hi,
In current implementation, we would expect all sources in identical resolutions. If 4 sources are rotated(such as 1920x1080 into 1080x1920), it should work properly. But if you have one source in 1080x1920 and the other thee in 1920x1080, it may not work properly in nvstreammux and nvmultistreamtiler plugins.

One possible solution is to create two separate pipelines. One runs in 1080x1920 and the other runs in 1920x1080.

I see. I would recommend to address this issue in next releases since in many industrial applications, cameras are oriented at different angles.
Not only just 90 degree rotation, but individual rotation at all the angles.

Hi,
We have checke this and a possible solution on DeepStream SDK 6.0 is to use nvdspreprocess plugin. Please check README of the sample:

/opt/nvidia/deepstream/deepstream-6.0/sources/apps/sample_apps/deepstream-preprocess-test/

It supports the two items which may be applied to the 90-degree rotated source:

  • Per group custom preprocessing on ROIs provided
  • Prepares raw tensor for inferencing

Please check and give it a try.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.