AI-102 Question 147
Select 3You have trained a custom vision model in Azure Custom Vision and are evaluating its performance using key metrics. Which metrics should you prioritize to determine how well the model differentiates between similar objects and handles incorrect classifications during testing?
- A
Precision
- B
Recall
- C
F1 Score
- D
Mean Average Precision (mAP)
- E
Training loss
Show answer and explanation
Correct answers: A, B, C
Explanation
Precision, Recall, and F1 Score are key metrics for evaluating a custom vision model's ability to differentiate between classes and handle incorrect classifications. Precision assesses the model's ability to avoid false positives, while Recall examines its ability to avoid false negatives. F1 Score provides a balanced view that combines both Precision and Recall. These metrics are more relevant in this scenario than mAP or Training loss, which are focused on other aspects of model performance.
- A. Correct.
Precision measures the proportion of correctly identified positive instances out of all instances predicted as positive. It is critical to evaluate how well the model avoids false positives when differentiating between classes.
- B. Correct.
Recall evaluates the proportion of correctly identified positive instances out of all actual positive instances. It is essential to assess whether the model correctly identifies true positives and avoids missing important classifications.
- C. Correct.
F1 Score is the harmonic mean of Precision and Recall, providing a balanced metric that considers both false positives and false negatives. It is particularly useful when you want to evaluate the trade-off between Precision and Recall.
- D. Incorrect.
Mean Average Precision (mAP) is commonly used in object detection tasks, but it is not the primary metric for evaluating model differentiation or classification accuracy in general custom vision scenarios.
- E. Incorrect.
Training loss measures the error during training but does not directly indicate how well the model performs during evaluation. It is not a primary metric for evaluating classification performance post-training.