Open gtingstad94 opened 4 months ago
Hi @gtingstad94, the opencv-builder container outputs a tarball of opencv deb's, which I then download/extract/install in the opencv container. If you extract the aarch64-libs deb in there, it has contrib libraries in it which get installed (like the cuda ones)
Also I test during cv2.cuda during the builds, and it wouldn't pass if it wasn't installed? https://github.com/dusty-nv/jetson-containers/blob/master/packages/opencv/test.py
Also, it's opencv:4.8.1-builder
or opencv:4.5.0-builder
for the name of the builder containers (and you can list them all with ./build.sh --list-packages
or ./build.sh --show-packages opencv*
)
Hi @dusty-nv, thank you for getting back to me. I have done a bit more digging and realize I have made a mistake. The additional openCV packages do install with the jetson-containers opencv module and I was simply trying to access them incorrectly.
I appreciate your support!
I am trying to install opencv with cuda support and the additional cuda-enabled modules. I have been able to successfully install the opencv image and verify that cuda is enabled, but it seems that the additional opencv-contrib packages that leverage gpu acceleration are not baked into the default docker image? Upon installation, and when running
cv2.getBuildInformation()
, I see a long list of to-be-built packages.I see that the Dockerfile in opencv_builder includes a line to bundle in the additional opencv-contrib files, but as far as I have been able to tell, this configuration is not used during the "vanilla" installation of the opencv image. I have tried building with
./build.sh opencv_builder
, as recommended here, but I receive a KeyError: "couldn't find package: opencv_builder".Apologies if the answer is clear to some. I'm a fresher with Docker.