Question 23
Domain 5: Implement Computer Vision and Document Intelligence SolutionsYou train a Custom Vision object detection model. The evaluation shows Precision=0.92 but Recall=0.61. What does this mean and what should you do?
Correct answer: A
Explanation
Precision of 0.92 means most predicted defects are real, while recall of 0.61 means the model finds only 61% of actual defects, so it is missing many true cases. In object detection, low recall usually means you should add more training images for the missed defect types so the model learns to detect them better.
Why each option is right or wrong
A. The model rarely makes mistakes but misses many real defects; add more training images of missed defect types
Precision of 0.92 means that when the model predicts an object, it is correct 92% of the time, so false positives are relatively low; recall of 0.61 means it identifies only 61% of the actual objects, leaving 39% of true defects undetected. In Custom Vision object detection, that pattern points to missed examples rather than overprediction, so the remedy is to add more labeled training images for the defect classes and scenarios the model is failing to detect.
B. The model detects most defects but many are false positives; increase the confidence threshold
C. The model is overfitting; reduce the number of training images
D. The model accuracy is acceptable; no action needed