Overview
Azure AI Vision is Microsoft’s prebuilt computer vision service that enables applications to analyze images and videos without requiring machine learning expertise or custom model training. It provides REST APIs and SDKs that allow developers to easily extract visual insights such as objects, text, faces, and image descriptions.
For the AI-900 exam, you are expected to understand what Azure AI Vision can do, which problems it solves, and how it differs from custom vision solutions—not how to build or tune models.
What Is Azure AI Vision?
Azure AI Vision is part of Azure AI Services and offers ready-to-use computer vision capabilities, including:
- Image analysis
- Optical Character Recognition (OCR)
- Facial detection and analysis
- Object detection
- Image tagging and categorization
These capabilities are powered by Microsoft-trained deep learning models and are accessed via APIs.
Core Capabilities of Azure AI Vision
1. Image Analysis
Azure AI Vision can analyze images to extract high-level insights, such as:
- Objects present in an image (for example, car, building, person)
- Scene descriptions in natural language
- Image tags and categories
- Visual features such as color distribution
Example use cases:
- Auto-generating image captions
- Content moderation
- Organizing image libraries
👉 Exam tip: Image analysis describes what is in an image, not where every object is located with precision.
2. Object Detection
Object detection identifies specific objects in an image and returns:
- Object names
- Bounding box coordinates
- Confidence scores
Example use cases:
- Detecting vehicles in traffic images
- Identifying products on store shelves
👉 Exam tip: Object detection includes location + object type, unlike image classification which only labels the image as a whole.
3. Optical Character Recognition (OCR)
OCR extracts printed and handwritten text from images and documents.
Azure AI Vision OCR supports:
- Multiple languages
- Structured and unstructured text
- Images, screenshots, and scanned documents
Example use cases:
- Digitizing receipts
- Reading license plates
- Extracting text from scanned forms
👉 Exam tip: OCR is about reading text, not understanding its meaning.
4. Facial Detection and Facial Analysis
Azure AI Vision can detect human faces in images and analyze non-identifying facial attributes, such as:
- Face location (bounding boxes)
- Facial landmarks
- Estimated age range
- Facial expressions
- Accessories (glasses, masks)
⚠️ It does NOT identify individuals.
Example use cases:
- Blurring faces for privacy
- Counting people in images
- Analyzing expressions in photos
👉 Exam tip:
- Facial detection = where faces are
- Facial analysis = attributes of faces
- Facial recognition = identity (not required for AI-900)
5. Image Tagging and Categorization
Azure AI Vision automatically assigns tags and categories to images, such as:
- “outdoor”
- “food”
- “animal”
These tags help with searchability and organization.
Example use cases:
- Image indexing
- Content filtering
- Metadata enrichment
👉 Exam tip: Tagging helps describe images at a high level, not detect precise objects.
Azure AI Vision vs Custom Vision
| Feature | Azure AI Vision | Azure Custom Vision |
|---|---|---|
| Prebuilt models | ✅ Yes | ❌ No |
| Requires training | ❌ No | ✅ Yes |
| Quick setup | ✅ Yes | ❌ No |
| Specialized scenarios | ❌ Limited | ✅ Strong |
| AI-900 focus | ✅ Yes | ⚠️ Limited |
👉 Exam takeaway:
If the question mentions no training, quick setup, or prebuilt models, Azure AI Vision is usually the right answer.
Responsible AI Considerations
Because Azure AI Vision can analyze images of people, Microsoft emphasizes:
- Privacy and security of image data
- Transparency in how visual data is processed
- Fairness and bias mitigation
- Appropriate use of facial analysis
👉 Exam tip: Facial capabilities often pair with Responsible AI principles in exam questions.
Common AI-900 Exam Scenarios
You should recognize Azure AI Vision when the scenario involves:
- Analyzing images without training a model
- Extracting text from images
- Detecting faces but not identifying people
- Automatically tagging or describing images
Key Exam Takeaways
- Azure AI Vision is a prebuilt computer vision service
- No machine learning expertise required
- Supports image analysis, OCR, object detection, and facial analysis
- Focuses on insight extraction, not identity
- Frequently tested in scenario-based questions
Go to the Practice Exam Questions for this topic.
Go to the AI-900 Exam Prep Hub main page.
The Data Community 