Domain 3 of 5 · Chapter 2 of 2

Azure Vision Services

Unlock the complete study guide + 1,040 practice questions across 16 full exams.

Bundled into the existing Microsoft Azure AI Fundamentals premium course — no separate purchase.

Included in this chapter:

  • The four Azure vision services and what each one does
  • Which service for which task
  • Exam-pattern recognition and distractor traps

Azure computer-vision services: best-for and prebuilt vs custom

ServiceBest forPrebuilt or custom
Azure AI VisionGeneral image analysis (objects, tags, captions) and OCR/Read for printed and handwritten textPrebuilt (no training required)
Azure AI Document IntelligenceExtracting fields, key-value pairs, and tables from invoices, receipts, IDs, and formsPrebuilt models plus optional custom-trained models
Azure AI Custom VisionImage classification and object detection on your own domain-specific labeled imagesCustom (you upload and label training data)
Azure AI FaceFace detection, facial analysis, and identity recognition or verificationPrebuilt (some recognition features under Limited Access)

Decision tree

Train on your own labeled images? Yes Azure AI Custom Vision classify / detect with your labels No Detect / recognize human faces? Yes Azure AI Face detect, analyze, verify identity No Extract fields from forms / invoices? Yes Azure AI Document Intelligence key-value pairs, tables, fields No Azure AI Vision prebuilt image analysis + OCR (Read) Default: prebuilt, no training. Custom labels, faces, or form fields branch to the dedicated service.
Cheat sheet

Cheat sheet

  • Azure AI Vision is the prebuilt image-analysis and OCR service
  • Image Analysis detects objects, tags, captions, and adult content
  • OCR via the Read API extracts printed and handwritten text
  • Document Intelligence extracts structured fields, not just raw text
  • Document Intelligence is the renamed Form Recognizer
  • Document Intelligence ships prebuilt models plus trainable custom ones
  • Custom Vision trains an image model on your own labeled images
  • Custom Vision does image classification or object detection
  • Start a Custom Vision model with about 50 images per label
  • Azure AI Face handles detection, verification, and identification
  • Face recognition is gated behind Limited Access
  • Face presence is Vision's job; face identity is Face's job
  • Prebuilt vs. custom is the top AI-900 vision selection rule
  • Invoices and receipts go to Document Intelligence, not Custom Vision
  • Liveness detection is Face's anti-spoofing presence check
  • Face Verify is one-to-one; Identify is one-to-many
  • Face detection must run before any other Face operation
  • Face emotion and gender detection were retired under Responsible AI
  • Read OCR returns words with bounding boxes and confidence scores
  • Run Vision OCR on-premises with the Read Docker container
  • Only a compact-domain Custom Vision model can be exported
  • Custom Vision needs separate training and prediction resources
  • Multiclass assigns one tag; multilabel assigns many
  • Custom template suits fixed layouts; custom neural handles varied ones
  • The layout model returns structure: text, tables, selection marks
  • Match the task to the right Image Analysis feature

Unlock with Premium — includes all practice exams and the complete study guide.

References

  1. Azure AI Vision overview
  2. OCR overview
  3. Document Intelligence overview
  4. Custom Vision overview
  5. Azure AI Face overview