Implement Custom Vision Models AI-102 Practice Questions (Page 7)

Questions 31–35

1. 31foundation · easy

   When publishing a custom vision model iteration, what information is required?

   • AA Docker container image and a container registry
   • BAn Azure Function and a storage account
   • CA training dataset and a test dataset
   • DA model name and a prediction resource
   Reveal Solution Discussion Select an answer first
2. 32expert · hard

   A developer wants to train a custom vision model using code and then publish it for consumption. They are using Python. Which sequence of SDK calls is correct?

   • APublish → create project → add tags → upload images → train
   • BUpload images → create project → train → publish → add tags
   • CCreate project → add tags → upload images → train → publish
   • DCreate project → train → add tags → upload images → publish
   Reveal Solution Discussion Select an answer first
4. 33application · medium

   After training a custom image classification model to identify defective vs. non-defective parts, the evaluation shows a precision of 0.90 and a recall of 0.60 for the 'defective' class. The production line manager says that missing a defective part is very costly, but false alarms are only a minor inconvenience. Which metric should the team focus on improving, and what is the most direct way to do so?

   • ASwitch to object detection to get bounding boxes for each defective part.
   • BImprove accuracy by balancing the dataset with more non-defective images.
   • CImprove precision by adding more images of non-defective parts.
   • DImprove recall by adding more images of defective parts and retraining the model.
   Reveal Solution Discussion Select an answer first
5. 34foundation · easy

   What is the purpose of publishing an iteration of a custom vision model?

   • ATo make the model available for prediction through a REST endpoint
   • BTo store the model in the Azure Blob Storage
   • CTo delete the training images from the project
   • DTo create a new project in Custom Vision
   Reveal Solution Discussion Select an answer first
6. 35application · medium

   A wildlife conservation group wants to use custom vision to count the number of elephants in aerial photos. They need to know how many elephants are in each image and where they are located. Which model type and labeling approach should they use?

   • AImage classification; label each image as 'elephant' or 'no elephant'.
   • BObject detection; label the entire image as 'elephant' if any elephant is present.
   • CImage classification; label each image with the number of elephants in the image.
   • DObject detection; label each elephant with a bounding box.
   Reveal Solution Discussion Select an answer first