AI-102 Question 152
Select 2You have trained a custom vision model using the Azure Custom Vision service. You need to publish the model to make it accessible for prediction requests. Which of the following steps must you perform?
- A
Assign the model to a prediction resource in the Custom Vision portal
- B
Export the model to use it in an offline scenario
- C
Publish the model using a specific iteration name
- D
Set up an Azure API Management instance to expose the model endpoint
- E
Use the Custom Vision REST API to enable the model for public predictions
Show answer and explanation
Correct answers: A, C
Explanation
To publish a Custom Vision model for online predictions, you must assign it to a prediction resource and explicitly publish it with a specific iteration name. These steps make the model available via an endpoint for prediction requests. Additional configurations like API Management are optional and not part of the core publishing process.
- A. Correct.
Assigning the model to a prediction resource is required to link the trained model with the Azure resource that will host the endpoint for prediction requests.
- B. Incorrect.
Exporting the model is only necessary for offline or edge deployment scenarios, not for publishing it online.
- C. Correct.
Publishing the model with a specific iteration name is a required step to make it accessible via the endpoint.
- D. Incorrect.
Setting up Azure API Management is not required to publish the model, though it can be used to manage API requests.
- E. Incorrect.
The Custom Vision REST API does not directly enable the model for predictions; publishing the model in the portal or via SDK is sufficient.