AI-102 Question 136
Select 3You are working on an Azure AI solution to automate the labeling of images for a machine learning model. You decide to use Azure Custom Vision for this task. Which steps are required to set up and label images effectively in Azure Custom Vision?
- A
Create a Custom Vision project and select the appropriate project type (e.g., classification or object detection).
- B
Upload all your images and rely on Azure to automatically label them without any intervention.
- C
Manually tag the uploaded images with appropriate labels or use pre-trained models for assisted labeling.
- D
Train the Custom Vision model after labeling the images to generate predictions.
- E
Deploy the model without training it, as the labeling process alone is sufficient for predictions.
Show answer and explanation
Correct answers: A, C, D
Explanation
To effectively label images using Azure Custom Vision, you must start by creating a project and selecting the appropriate project type. After uploading images, you need to label them manually or use assisted labeling. Once labeling is complete, the model must be trained to learn from the data and generate predictions. Deploying the model without training it will not yield a functional AI solution.
- A. Correct.
Correct. Creating a Custom Vision project and selecting the right project type (classification or object detection) is a necessary first step for labeling images and training a model.
- B. Incorrect.
Incorrect. Azure Custom Vision does not automatically label images without intervention. You need to manually label images or use assisted labeling to ensure accurate model training.
- C. Correct.
Correct. Manually tagging the images or using assisted labeling helps provide the model with the necessary data to learn and make accurate predictions.
- D. Correct.
Correct. Training the model after labeling is essential to allow the model to learn from the provided data and generate predictions.
- E. Incorrect.
Incorrect. Deploying the model without training it will not work, as labeling alone does not create a functional model capable of predictions.