AI-900: How to Improve an Underperforming Image Classification Model?

Learn how adding more images to the training set is the most effective way to boost the performance and accuracy of an image classification model. Prepare for the AI-900 Azure AI Fundamentals certification exam.

Question

You are training an image classification model, but it achieves less than satisfactory evaluation metrics. How might you improve it?

A. Reduce the size of the images used to train the model.
B. Add a new label for “unknown” classes.
C. Add more images to the training set.

Answer

If your image classification model is not achieving satisfactory evaluation metrics, the best approach is to add more images to the training set (Option C).

C. Add more images to the training set.

Explanation

Here’s why:

1. More training data helps the model learn a wider variety of features and patterns associated with each class. With additional examples, the model can better generalize to new, unseen images.
2. A larger training set reduces the risk of overfitting, where the model becomes too specialized to the limited data it was trained on. More diverse training images help the model become more robust.
3. Adding more images is particularly helpful if some classes have significantly fewer training samples than others (class imbalance). Increasing the number of underrepresented class images helps the model learn those classes more effectively.

In contrast, reducing image size (Option A) can lead to loss of important details and will likely worsen performance. Adding an “unknown” label (Option B) doesn’t address the root issue of insufficient training data for the existing classes.

To summarize, expanding the training dataset with additional, diverse images is the most reliable way to improve an underperforming image classification model. A richer training set enables the model to learn better feature representations and generalize more effectively to real-world data.

Adding more images to the training set can improve the model’s performance by giving it more data to learn from, which can lead to better generalization and more accurate predictions. More diverse and representative images can enhance the model’s evaluation metrics.

Microsoft Azure AI Fundamentals AI-900 certification exam practice question and answer (Q&A) dump with detail explanation and reference available free, helpful to pass the Microsoft Azure AI Fundamentals AI-900 exam and earn Microsoft Azure AI Fundamentals AI-900 certification.