Implement Custom Vision Models AI-102 Practice Questions (Page 5)

Questions 21–25

1. 21application · medium

   A retail company wants to train a custom vision model to detect whether store shelves are fully stocked or have empty spots. They have 1,500 images of shelves. They plan to use Azure Custom Vision. What should they do to train the model?

   • AUpload images and train a multi-label classification model with tags for each product
   • BUpload images and let Custom Vision automatically label them
   • CUpload images and train an object detection model without tagging
   • DUpload images, tag them as 'stocked' or 'empty', and train an image classification model
   Reveal Solution Discussion Select an answer first
2. 22foundation · easy

   In a code-first approach to creating a custom vision model, what is the first step after creating the Custom Vision client?

   • APublish the model
   • BCreate a new project
   • CUpload images
   • DTrain the model
   Reveal Solution Discussion Select an answer first
4. 23application · medium

   A team is labeling images for an object detection model that identifies different types of vehicles in traffic camera images. What labeling approach is required?

   • AOnly label images that contain a single vehicle
   • BApply a single tag to the entire image indicating the most common vehicle
   • CDraw a bounding box around each vehicle and assign the correct class tag
   • DTag the image with all vehicle types present without boxes
   Reveal Solution Discussion Select an answer first
5. 24foundation · easy

   An application needs to send an image to a published custom vision model and receive a prediction. What is the primary way to do this?

   • AExport the model as a TensorFlow model and run it locally
   • BCall the prediction endpoint using the REST API or an SDK
   • CUse Azure Blob Storage to store the image and then manually review it
   • DUpload the image to the Custom Vision portal and view the results there
   Reveal Solution Discussion Select an answer first
6. 25application · medium

   A retail company has trained a custom vision model to identify products on store shelves. The model is currently in the training portal and performs well on the test set. The development team wants to integrate the model into a mobile app that store associates will use to scan shelves. The app needs to send images to the model and receive predictions. What should the team do to make the model accessible to the app?

   • AExport the model as a TensorFlow model and embed it directly into the mobile app binary.
   • BCreate a new project in the Custom Vision portal and upload the app's images for real-time classification.
   • CUse the training API endpoint to send images and get predictions.
   • DPublish the model to an Azure Container Instance and use the prediction endpoint URL and key in the app.
   Reveal Solution Discussion Select an answer first