Deploy a Clea Studio AI workflow to SOM-SMARC-Dragonwing-IQ8
Overview
This procedure deploys a Clea Studio AI project to SOM-SMARC-Dragonwing-IQ8 through Clea.
The reference hardware is SOM-SMARC-Dragonwing-IQ8 installed on the SECO EVAL-KIT-SOM-SMARC with Carrier-SMARC-E71.
The example uses the Face Detection on Camera project. Studio AI creates the edge application release, Clea makes the release available to the target, and Clea Edgehog provides the device-level deployment and runtime view.
What you need in advance
Prepare:
- access to Clea Studio AI;
- access to the Clea Portal;
- SOM-SMARC-Dragonwing-IQ8 installed on EVAL-KIT-SOM-SMARC;
- Clea OS running on the target;
- the IQ8 evaluation device registered in Clea Edgehog;
- network connectivity between the target and Clea;
- the Face Detection on Camera Studio AI project;
- any input file or camera required by the configured Studio AI workflow.
Before starting the deployment, open:
Clea Edgehog > Fleet
Select the IQ8 evaluation device and confirm that its connection state is Connected.
1. Open the Studio AI project
Open Clea Studio AI and select:
Face Detection on Camera
The reference project contains:
StaticVideoInput > Face Detection HMI > Image logger
Open each block and check its configuration.
For the reference preview, StaticVideoInput uses:
Source: video_face_detection.mp4
Drop policy: target-fps
Drop policy value: 15
Frame width: 320
Frame height: 240
Run the preview and confirm that the face-detection output is correct before creating an edge release.
Save the project.
2. Create the Clea edge release from Studio AI
From the project editor, select:
Deploy to edge
The deployment dialog opens.
Enter an Application name.
Use:
Face Detection on Camera
Enter a Release name.
Use a versioned value that can be identified later in Clea, for example:
1.0.0
Click:
Deploy
Wait for Studio AI to complete the release creation.
The application and its release are now available to the Clea deployment flow.
3. Open the IQ8 device Applications page
Open the Clea Portal.
Go to:
Clea Edgehog > Fleet > <IQ8 evaluation device>
Open the device tab:
Applications
The page contains an Install Applications section and a Deployed Applications list.
In Install Applications, select:
Face Detection on Camera
Select the release created in Studio AI, for example:
1.0.0
Start the installation.
Wait for Clea to create the application deployment on the IQ8 target.
4. Open the deployed application
Stay on the device Applications tab.
Under Deployed Applications, open:
Face Detection on Camera
The application detail page shows the release deployed on the device.
The generated edge application is organized as containers. Expand the container entries and inspect:
- image;
- device mappings;
- volumes;
- networks.
For the reference face-detection workflow, the deployment represents the input, inference, and output stages created in Studio AI.
Check the device mappings carefully when the application uses a physical camera or another hardware interface. The mapping has to expose the corresponding Linux device to the container that consumes it.
5. Start the application
Use the start control in the application detail page.
Watch the application state while the containers start.
Each required container should reach the expected running state.
If a container stops during startup, open its details and check the runtime information available in Clea. Typical causes include:
- missing device mapping;
- unavailable input source;
- missing model asset;
- volume path mismatch;
- container image pull failure;
- network access required by the application.
Resolve the reported runtime condition, then start the application again.
6. Validate the face-detection workflow
Use the same input configuration that was checked in the Studio AI preview.
For the reference project, confirm that the packaged video source can be read and processed on the IQ8 target.
Check:
- input initialization;
- face-detection container state;
- output from
Image logger; - container stability;
- resource usage during inference.
When the project is adapted to a live camera, check the camera device mapping on the E71 platform before starting the deployment. Record the Linux device path and capture format used by the application.
7. Check the deployment from the device page
Return to:
Clea Edgehog > Fleet > <IQ8 evaluation device> > Applications
Confirm that Face Detection on Camera appears under Deployed Applications with the expected release.
Open the application again and verify the container configuration.
The application detail page is the main Clea view for checking how the Studio AI flow has been translated into the runtime deployed on the device.
8. Create a new release after a Studio AI change
Return to Studio AI and modify one project setting, such as an input parameter or model configuration.
Save the project.
Open:
Deploy to edge
Keep the application name:
Face Detection on Camera
Create a new Release name, for example:
1.0.1
Click Deploy.
In Clea, return to:
Clea Edgehog > Fleet > <IQ8 evaluation device> > Applications
Select the new release and install it on the evaluation device.
Keeping a new release identifier for each Studio AI change makes it easier to compare deployments and return to a previous application configuration during validation.
Result
The Studio AI workflow has been packaged as a Clea application release and deployed to SOM-SMARC-Dragonwing-IQ8 on the E71 evaluation kit.
The complete path is:
Clea Studio AI
|
| Deploy to edge
v
Clea application + release
|
| Clea Edgehog > Fleet > device > Applications
v
SOM-SMARC-Dragonwing-IQ8 on EVAL-KIT-SOM-SMARC
This path is useful when the AI workflow can be deployed directly as a Clea-managed edge application. The Arduino App Lab route remains available for applications that use the Studio AI component inside a broader Arduino application.
Compatibility
The same Clea deployment pattern can be used with Qualcomm-based products supported by Clea OS. Review the hardware-specific device mappings, BSP support, camera interfaces, and accelerator access before deploying the application to another target.