You can find the image on docker hub.
If you need a docker image like this one but not for jetson, this repo has a twin!
Feel free to open an issue if you are having some problems.
Nvidia ecosystem can be quite difficult, I know.
Note: This image works properly only on Jetson nano, TX2, AGX Xavier, Xavier NX, (not yet but soon: AGX Orin, Orin NX, Orin nano)
- l4t-base:r32.7.1 (Jetpack 4.6.x)
- Python 3.6
- Opencv 4.8.0
- Numpy
Details:
- cuda version: 10.2
- cudnn version: 8.2
For details on opencv build you can check the build script.
Base Images:
- Build: made on jetson without docker
- Runtime: nvcr.io/nvidia/l4t-base:r32.7.1
If you are familiar with the jetson environment you are probably asking why I made this image and not used dustynv/opencv
The reasons are:
- I have to use opencv version 4.8.0 (in dustynv/opencv opencv version is 4.5.0)
- I wanted to have more control on the image
Before running the docker image on jetson device, make sure that you have docker runtime nvidia inplace:
-
Open:
sudo nano /etc/docker/daemon.json -
Edit from:
{ "runtimes": { "nvidia": { "path": "nvidia-container-runtime", "runtimeArgs": [] } } }To:
{ "runtimes": { "nvidia": { "path": "nvidia-container-runtime", "runtimeArgs": [] } }, "default-runtime": "nvidia" } -
Restart docker:
sudo systemctl restart docker
Then add user to docker group
sudo usermod -aG docker $USERsudo reboot
See docker-compose.
To run with window: from jetson:
- export DISPLAY=:0
- xhost +
- go to X/build_opencv where X is whatever origin folder
docker build .- Once finished (on jetson nano it will take 7h) run the docker with a shared folder (to get the build files)
docker run -v HOST/X/installer:/shared -i -t tag bash - Copy built packages from docker to host machine
cp OpenCV-4.8.0-aarch64.* /shared/