AI-102 Question 151
Select 2You have trained a custom image classification model using Azure Custom Vision. Now, you need to publish it to make it accessible for integration into a web application. Which steps are required to successfully publish the model?
- A
Assign the model to a prediction resource in the Azure portal.
- B
Use the Azure CLI to deploy the model to Azure Functions.
- C
Click on the 'Publish' button in the Custom Vision portal and provide a name for the model.
- D
Enable the prediction endpoint in the Azure Custom Vision portal.
- E
Export the model as a TensorFlow file and deploy it to a Kubernetes cluster.
Show answer and explanation
Correct answers: A, C
Explanation
Publishing a custom vision model requires assigning it to a prediction resource and using the 'Publish' button in the Custom Vision portal to make the model accessible via a REST API endpoint. Steps like exporting the model or deploying it to Kubernetes are not required for simple publishing within the Azure ecosystem.
- A. Correct.
Correct: Assigning the model to a prediction resource links it to an Azure resource that provides the necessary infrastructure for hosting and serving predictions.
- B. Incorrect.
Incorrect: The Azure CLI is not directly used to publish models in Custom Vision. Publishing is done via the Custom Vision portal.
- C. Correct.
Correct: Clicking 'Publish' in the Custom Vision portal and providing a name is a required step to make the model accessible via the prediction endpoint.
- D. Incorrect.
Incorrect: The prediction endpoint is automatically enabled when you publish the model; there is no separate action required to enable it.
- E. Incorrect.
Incorrect: Exporting the model is an optional step for offline or specific deployment scenarios, but it is not required for publishing the model in Azure Custom Vision.