5.1. Analyze Images

💡 First Principle: Image analysis is about extracting meaning from pixels. Azure AI Vision provides pre-trained models that understand image content at multiple levels—from simple tags ("dog", "outdoor") to rich captions describing the scene. The exam tests which visual feature to request based on your scenario.

What breaks without proper feature selection: Requesting all features wastes API calls and increases costs. Requesting the wrong feature gives you data you can't use. Understanding the output format of each feature is essential for parsing results correctly.

Feature selection framework:

• Need a description? → Caption or Dense Captions
• Need searchable keywords? → Tags
• Need to find specific objects? → Object Detection
• Need to extract text? → Read API (OCR)
• Need content moderation? → Adult Content detection