Closed stvogel closed 4 years ago
So I dug deeper into this and I saw that the cmake for opencv did not create a Makefile due to errors.
I got the somehow famous error about #include <sys/videoio.h> no such file
and all solutions found on the internet didn't work out.
I just set CTO_CUDA_BUILD to an empty string in the Makefile, so opencv is not build with CUDA and than everything runs fine.
Thanks for making and sharing this repo.
Thanks for providing these docker-images. I hoped getting around compiling opencv ;-) I used the docker image jetsonnano-cuda_tensorflow_opencv:10.0_2.1.0_4.3.0-20200515
But when doing
import cv2
in python3.6 I get the errorlibnppc.so.10 not found
. I assume the base-nvidia docker has upgraded to jetpack 4.4 which uses cuda 10.2 (instead of jetpack 4.3 which uses cuda 10.0)Might I be right? Is the solution to use another base-container in Makefile/Dockerfile: instead of nvcr.io/nvidia/deepstream-l4t:4.0.2-19.12-base resp nvcr.io/nvidia/l4t-base:r32.3.1?
I cannot figure out what to use instead as the nvcr-catalog is ... well I cannot find anything there :-(
Any help is appreciated. Thanks