NVIDIA Jetson Orin NX

Developer Preview of deploying the EyePop.ai runtime to the Jetson Orin NX

Jetson Orin Runtime

Run the EyePop runtime on an NVIDIA Jetson Orin with Docker.

Requirements

• NVIDIA Jetson Orin with JetPack installed

• Docker

• NVIDIA container runtime

• EyePop dashboard access

• Python3.12 installed on Jetson Orin NX

1. Get The Server Files From The Dashboard

Open https://dashboard.eyepop.ai, go to My Servers, and click Add Server.

The dashboard provides:

• eyepop-instance.yml

• a Docker registry login command

Save eyepop-instance.yml on the Jetson, for example in ~/eyepop/eyepop-instance.yml.

The file should look similar to this:

eyepop-account-uuid: <your EyePop account UUID>
eyepop-api-key: <your EyePop API key>
instance-name: <your server name>

Use the dashboard-provided values as the source of truth.

Make the file readable by Docker:

chmod 644 eyepop-instance.yml

2. Log In To The Registry

Run the Docker login command shown in the dashboard on the Jetson.

It will look like this:

printf '%s' '<registry key secret>' | docker login registry.eyepop.ai \
  -u '<registry key username>' \
  --password-stdin

The registry key secret is sensitive. Do not paste it into support tickets or logs.

3. Start The Runtime

From the directory containing eyepop-instance.yml, run:

docker run -d \
  --name eyepop-runtime \
  --restart unless-stopped \
  --runtime nvidia \
  -p 8080:8080 \
  -e NVIDIA_VISIBLE_DEVICES=all \
  -e NVIDIA_DRIVER_CAPABILITIES=compute,utility \
  -v "$PWD/eyepop-instance.yml:/etc/eyepop-instance.yml:ro" \
  registry.eyepop.ai/ai/runtime-cuda-jetson:latest

Use the Jetson runtime image:

registry.eyepop.ai/ai/runtime-cuda-jetson:<tag>

Do not mount host CUDA libraries into the container, and do not set NVIDIA_DISABLE_REQUIRE.

4. Verify

Check health:

curl http://127.0.0.1:8080/health

Check readiness:

curl http://127.0.0.1:8080/ready

Follow logs:

docker logs -f eyepop-runtime

The first run can take a minute while the runtime registers. The first model run can also take longer while the model is downloaded and cached.

Local Webcam Access

Runtime release v3.46.0 adds local V4L2 webcam access.

To allow local webcam sources, add this to eyepop-instance.yml:

allow-v4l2: true

Mount the camera device by adding this flag to docker run:

--device /dev/video0:/dev/video0

Use a source URL like:

v4l2:///dev/video0

The camera device path can vary by system. Check available devices with:

ls /dev/video*

Stop Or Update

Stop the runtime:

docker stop eyepop-runtime
docker rm eyepop-runtime

Update to a newer runtime tag:

docker pull registry.eyepop.ai/ai/runtime-cuda-jetson:<tag>
docker stop eyepop-runtime
docker rm eyepop-runtime

Then run the start command again with the new tag.

Install Python 3.12 from the deadsnakes PPA

Install the prerequisite for adding custom PPAs,

apt install software-properties-common -y

Add deadsnakes PPA,

add-apt-repository ppa:deadsnakes/ppa

Update the package lists to include packages from the newly added PPA,

apt update -y

Now, you can install Python 3.12 using the apt package manager.

apt install python3.12

Verify Installation,

python3.12 --version

Output:

root@vps:~# python3.12 --version
Python 3.12.0
root@vps:~# 

Once installed Python 3.12 using the APT package manager, the PIP will not be installed by default. Install PIP, run the following command,

curl -sS https://bootstrap.pypa.io/get-pip.py | python3.12

Now check the PIP version using the below command,

pip3.12 -V

Output:

root@vps:~# pip3.12 -V
pip 24.0 from /usr/local/lib/python3.12/dist-packages/pip (python 3.12)
root@vps:~#