AI-102 Question 141
Select 3You are tasked with building a custom image classification model to identify different types of fruits in an e-commerce application. You decide to use Azure Custom Vision for this purpose. After uploading the images and tagging them appropriately, what steps should you take to ensure the model is trained correctly and performs well?
- A
Split the dataset into training and testing sets before starting the training process.
- B
Use a pre-trained model in Azure Custom Vision to speed up training and improve accuracy.
- C
Manually adjust the learning rate and batch size in Azure Custom Vision to optimize performance.
- D
Test the model using images it has not seen before to evaluate its accuracy.
- E
Export the model to Azure Machine Learning for additional hyperparameter tuning.
Show answer and explanation
Correct answers: A, B, D
Explanation
To successfully train and evaluate a custom image classification model in Azure Custom Vision, it is essential to split the dataset into training and testing sets, take advantage of pre-trained models to enhance performance, and evaluate the model using unseen data to ensure it generalizes well. Azure Custom Vision abstracts complex hyperparameter tuning, so manual adjustments like learning rate and batch size are not necessary for this service.
- A. Correct.
Splitting the dataset into training and testing sets is essential to evaluate the model's accuracy on unseen data and prevent overfitting.
- B. Correct.
Azure Custom Vision provides the option to use pre-trained models, which can significantly speed up training and improve accuracy, especially for image classification tasks.
- C. Incorrect.
Azure Custom Vision does not allow manual adjustment of learning rate or batch size, as it abstracts these complexities for users.
- D. Correct.
Testing the model on unseen data is a standard practice to evaluate its generalization performance and ensure it works well in real-world scenarios.
- E. Incorrect.
While exporting a model to Azure Machine Learning is possible, it is not required for basic training and evaluation of a custom image classification model in Azure Custom Vision.