Not able to detect CUDA enabled device in Opencv GPU

Question

Hello,

I am trying to use Opencv GPU with CUDA.I ahve used the CMake for compiling the opencv 2.3.1 with cuda 4.0 But when I am trying to use the function cv::gpu:getCudaEnabledDevice() it returns me zero means no device available.

Though I have CUDA enabled gpu in my system.Anybody please help me with this I have tried a lot,In cmake it shows that no CUDA_SDK_ROOT_CIR not found.Which path should be given here?

Thanks in advance,

Answer 1

您应该从此处将NVIDIA驱动程序更新为最新版本

Answer 2

Are you sure that your cuda enabled gpu is not indexed as device zero? Also, try executing cuda SDK examples to check correct cuda setup.

Answer 3

Follow the checklist:

1. Check your video propietary driver version
2. Check the binary you have compiled with ldd - are the libraries on place?
3. In CUDA samples compile and run deviceQuery - check the output
4. Use the device in dedicated mode, ie if you use X11, check the config that any reference to nvidia driver and so forth are completely wiped out from here.
5. Compile X11 and supplementary libraries from source, DO NOT obtain it from pre-compiled binary packages

This checklist covers 80% of CUDA mailfunctions for sure - checked it myself.