No CUDA-capable device is detected in function 'getDevice' jetson xavier

Robotics & Edge Computing Jetson & Embedded Systems Jetson AGX Xavier
,
1

Hi, I tried to run my python application on Jetson AGX Xavier.
The python application runs in a docker container and uses opencv. The docker image is based on the l4t_base image.
I compiled opencv with cuda enabled in the container and if I run print(cv2.getBuildInformation())
it shows

NVIDIA CUDA: YES (ver 10.2, CUFFT CUBLAS FAST_MATH)
NVIDIA GPU arch: 72
NVIDIA PTX archs:
cuDNN: YES (ver 8.0)

But when I run my python code with cv2.cuda.cvtColor
it shows

signal: aborted (core dumped)
terminate called after throwing an instance of ‘cv::Exception’
what(): OpenCV(4.4.0) /tmp/build_opencv/opencv/modules/core/src/cuda_info.cpp:84: error: (-217:Gpu API call) no CUDA-capable device is detected in function ‘getDevice’

I also ran the following in both the container and the host machine.

import cv2
count = cv2.cuda.getCudaEnabledDeviceCount()
print(count)

On the host machine, it returns 1 but in the container, it returns 0.

How should I solve this?
Thanks

5

Sorry for the late response, is this still an issue to support?

Thanks

6

Hi,

The default OpenCV version doesn’t enable GPU support.
Please build it from source to get the CUDA support.

Here is a building script for your reference:

Thanks.

7

Hi,
I built opencv from source with cuda enabled but I cannot get the device. Should I rebuild opencv using your script?

8

Hi,

YES. Please give it a try.

Thanks.