Correct! Multi-class classification assigns exactly one label per image from a set of mutually exclusive categories, while multi-label classification can assign multiple labels to the same image since categories aren't mutually exclusive.
Incorrect. Multi-class assigns one label per image from mutually exclusive categories, while multi-label can assign multiple labels to the same image since categories aren't mutually exclusive.
Excellent! Background or negative examples teach the model what doesn't belong to any of your target categories, helping it avoid false positives by distinguishing meaningful features from background elements.
Incorrect. Background examples help the model learn what doesn't belong to your target categories, preventing it from confusing irrelevant background elements with meaningful features.
Correct! Publishing the model makes it available via the prediction API, which is required for making predictions through Python code or any external application.
Incorrect. You must publish the model from the Azure portal to make it available via the prediction API for use in Python code or external applications.
Excellent! AutoML supports advanced computer vision tasks including segmentation models when you provide ground truth masks, expanding beyond the classification and object detection capabilities of Custom Vision.
Incorrect. AutoML supports segmentation models with ground truth masks, demonstrating its advanced computer vision capabilities beyond what Custom Vision offers.
Perfect! The probability threshold acts as a confidence filter - only predictions with confidence scores above this threshold are returned, helping you control the trade-off between precision and recall.
Incorrect. The probability threshold determines whether predictions are returned based on their confidence scores - only predictions above the threshold are shown.
Complete the quiz to see your results.