Skip to main content

Model Deployment

There are two types of model deployment:

  • Code Deployment: Use the Geti™ SDK to deploy models locally or integrate them into Python-based applications.
  • OpenVINO™ Model Server Deployment: Deploy models on a server using OpenVINO™ Model Server, enabling inference over a network via an HTTP interface.

Deploy vs download the model​

While you can download all types of models, deployment is only available for OpenVINO™ models. The key difference is that a downloaded model contains only the model itself, whereas a deployment gives you executable code.

Let's download the model​

To download other types of models, navigate to the Models option in the sidebar and select the desired model. A tab with several model variants will appear. Choose one and download it by clicking the download icon download.

Downloading model

Let's deploy the model​

The Deployments page allows you to export deployment code for the trained models.

Once you have tested and optimized the model, you are ready to download and deploy your solution. To download the code, click on Select model for deployment. In the dialog box, choose the model architecture and its version as well as the model optimization if available. In the case of a task chain project, you will need to select two models through a wizard. Upon your selection, click Download. Additionally, you can choose between Code Deployment and OpenVINO™ Model Server Deployment.

Compatibility​

For Geti™ SDK versions up to 2.6, the SDK and platform versions must match to ensure compatibility. Starting with version 2.6, Geti™ and SDK releases are fully compatible across versions.

License​

License file can be found inside the deployment package. Models are provided under APACHE 2.0 license and they can be used without any limitation both online and offline.