I am trying to use the OpenCV ArUco module and accelerate using CUDA on the Jetson Xavier NX, in Python.
I have the module working well but I am showing no GPU usage.
Could anyone help me with this?
Edit: I know NVIDIA has done accelerated april tags using the ISAAC SDK, but I don’t have a machine now to run ISAAC so I can’t use that yet.
Can you point to the documentation that makes you expect the GPU to be exercised via CUDA for ArUco marker detection? Have you looked at the source code? Are you not instead hoping for OpenCL support under the hood?
@der_luftmensch thanks for the reply – it’s not that I am expecting the GPU to be utilized, but trying to figure out if it can be used and if anyone has experience implementing it for this application.
I am almost 100% sure that the ArUco does not have a CUDA implementation. If it does it should be accessible under the cv::cuda namespace. To confirm you should look at the source and see how OutputArray is dealt with internally.