Microsoft Azure AI Engineer Associate AI-102 Practice Question

The metric is mean Average Precision (mAP). For each class, Custom Vision computes Average Precision as the area under the precision-recall curve at the chosen IoU threshold (0.5 by default). It then averages these AP values across all classes, producing mAP. This single number balances precision and recall across classes and is the primary score used to compare object-detection iterations. F1 Score only combines precision and recall at one operating point, RMSE measures regression errors, and the Dice coefficient is mainly used for segmentation tasks, so none of these are the metric displayed by Custom Vision for object detection.