[Deepstream] How can I do my own image processing in C++?

[yuki.akimoto.sub]

Thanks for your reply !

I want to acquire images from a 360-degree camera using an image capture element such as nvv4l2camerasrc, crop the images around a specified viewing direction (ψ, θ) , send the cropped images to the pipeline, and draw them using nv3dsink.
The viewing direction (ψ, θ) is received by UDP communication and should be changed dynamically.

I considered using nvdewarper, but found that it is not possible to dynamically change the direction of the cutout, since it is specified by the parameter config-file.

Reference ↓

OTA update source config

I don’t understand what your code is for, but deepstream only support “on-the-fly” model update(On the Fly Model Update — DeepStream 6.1.1 Release documentation). You can not update other things on the fly.

Is it possible to do what I want to do?
I hope you will reply when you have time!