AI-102 Question 154
Select 3You have trained a custom vision model in Azure Custom Vision to classify images of damaged and undamaged vehicles. You now want to publish the model and make it accessible for scoring via an API. Which steps should you take to successfully publish the model?
- A
Assign the model to a specific iteration and publish it with a unique name.
- B
Export the model to an edge-compatible format such as ONNX or TensorFlow.
- C
Ensure the Custom Vision project is associated with a valid Azure Resource Group and Prediction Resource.
- D
Generate a prediction endpoint URL and an API key for your published model.
- E
Manually deploy the model to an Azure Kubernetes Service (AKS) cluster.
Show answer and explanation
Correct answers: A, C, D
Explanation
To publish a Custom Vision model, you must assign it to a specific iteration and provide a unique name. Additionally, the project must be linked to a valid Azure Resource Group and Prediction Resource. After publishing, a prediction endpoint URL and API key are automatically generated for using the model. Exporting the model or deploying it to AKS is not a prerequisite for publication; these are optional steps for specific use cases.
- A. Correct.
Correct. To publish a Custom Vision model, you must assign it to a specific iteration and provide a unique name for publication.
- B. Incorrect.
Incorrect. Exporting the model is not required for publishing it. Exporting is only necessary for on-premises or edge deployment scenarios.
- C. Correct.
Correct. The Custom Vision project must be linked to a valid Azure Resource Group and Prediction Resource to enable publishing and API access.
- D. Correct.
Correct. Once published, a prediction endpoint URL and an API key are generated to allow external applications to interact with the model.
- E. Incorrect.
Incorrect. Deploying the model to an Azure Kubernetes Service (AKS) cluster is an optional step for specific scenarios, such as scaling, but it is not required for publishing the model.