Deployment

Overview¶

To demonstrate the capabilities of AiCSD, the Intel team has created a sample AI pipeline using Intel Geti models and SDK.

With the Intel Geti platform, customers can train their own models for object detection, classification, segmentation, and anomaly. After the model has been trained, customers can export the models and deploy them seamlessly on AiCSD.

Learn more: Intel Geti

Deploy your Geti model¶

Once you export the trained model from Geti you can upload it to the AiCSD Web UI or add the model into the file system manually.

Web UI (recommended)¶

Navigate to the Upload Models tab.

Give the model a name, select Geti as the model type and upload the zip file exported from Geti using the Choose File button. Then click the Save button.

Manual¶

Create a subfolder under the models folder inside the AiCSD repo on the gateway machine. You can use any name, for example models/cells.

After that, unzip the exported models file from Geti and copy the deployment folder inside the models/cells folder in the AiCSD repo.

Create a task with new pipeline¶

To create a new task with the new deployed model, go to the task page on the AiCSD UI, and click on "Add Task".

On the "Pipeline" dropdown menu, it will automatically appear the new model you deployed.

Remaining Workflow¶

Now follow the rest of the Workflow as illustrated here

BSD 3-Clause License: See License.