4.3. Reflection Checkpoint: Computer Vision Mastery

Skipping this checkpoint means carrying confusion into the exam. Computer Vision questions appear throughout, and mixing up classification, detection, and OCR leads to wrong answers. Imagine seeing "identify objects with their positions" and not instantly recognizing that "positions" signals object detection.

Consider these questions like a final exam before moving on. For instance, the specialized domain models question trips up many people—only celebrities and landmarks are supported, not animals or cars. If you hesitate on any question, review Section 4.1 or 4.2 before proceeding.

1. Which computer vision capability returns bounding box coordinates?

   • Object detection. Image classification returns only a single label without position information.

2. A company needs to digitize medical records by extracting text from scanned documents. Which AI workload is this?

   • Computer Vision (specifically OCR). The input is images of documents.

3. What are the two specialized domain models supported by Azure AI Vision for image categorization?

   • Celebrities and Landmarks. Not animals, cars, or plants.

4. What additional information does an image description task return with each phrase?

   • Confidence score. This indicates how certain the model is about each description.

5. Which Azure AI Face feature helps identify whether a person is wearing glasses or headwear?

   • Face attributes. This is part of the Face Detect API's optional attribute analysis.