Hello,
I do install OpenCv via SdkManager however when I check OpenCv it says no to Compiled Cuda.
And also python3 does not see the cv2.
What is the problem and how to solve these?
Should I really use jetsonhacks’ OpenCv installation?
Why does not sdkmanager enable Compiled Cuda and python3 for OpenCv
The opencv from sdkmanager only provides necessary lib/functions other sdk components need. For enabling python and cuda, the steps have no difference between upstream openCV, so I think you could directly follow the installation steps from other users.
Seriously, why not just enable CUDA + Python3 support for OpenCV in the SDK?
If NVIDIA is actually trying to build an attractive embedded GPU ecosystem for devs, it’s a bit baffling that you’re leaving this unaddressed and relying on devs to yak shave before they can actually do proper dev work.
Especially when a cursory search of “cuda support opencv2” on this forum shows about 50+ threads since 2017 asking about this very same thing.
Many of us understandably expect that this feature exists, given that it’s such a commonly enabled cmake flag, and moreover we’re talking about CUDA support on an NVIDIA platform here.
This comment is quite legit indeed.