Source video crop before inference

• Hardware Platform = Jetson
• DeepStream Version = 4.0.2
• JetPack Version = 4.3-b134_arm64
• TensorRT Version = 6.0
** Source type = CSI Camera **

How can I crop the source video to a constant ROI in the deepstream-app source code before inference?
I tried using g_object_set to set “src_crop” and “dst_crop”, but it says that nvvideoconvert does not have such property.


I had a similar problem, help would be much appreciated.

Deepstream 4.0.2 is not supported any longer. Please update to the latest DeepStream 5.0 GA.
You can get it here: