Answer-first summary for fast verification
Answer: Label images manually, choose an image classification model, train with default settings, and evaluate using F1 score.
For image classification, manual labeling is often more accurate than automated methods. An image classification model is appropriate here as opposed to an object detection model, which is more suited for scenarios where the location of objects within the image is important. Training with default settings is a good starting point, and evaluating using the F1 score, which balances precision and recall, is a robust method for assessing classification models.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
You are tasked with developing a custom image classification model for a retail company to identify different types of products on their shelves. The company has provided you with a dataset of 10,000 images, divided into 50 categories. Describe the steps you would take to train this model using Azure AI Vision, including how you would label the images, the type of model you would choose, and how you would evaluate its performance.
A
Label images manually, choose an object detection model, train using default settings, and evaluate using accuracy.
B
Use automated labeling, choose an image classification model, train with custom hyperparameters, and evaluate using precision and recall.
C
Label images manually, choose an image classification model, train with default settings, and evaluate using F1 score.
D
Use automated labeling, choose an object detection model, train with custom hyperparameters, and evaluate using mean average precision.
No comments yet.