Study Notes

Appearance

Domain 3: Computer Vision Workloads on Azure ​

<- Previous Domain | Overview | Next Domain ->

Weight: 15-20%

This domain tests whether you can identify common computer vision scenarios and match them to Azure AI Vision and Azure AI Face capabilities.

Common Computer Vision Solutions ​

ScenarioWhat It MeansExample
Image classificationAssign one or more labels to an imageIdentify product category
Object detectionDetect object instances and locationsFind damaged items on a conveyor belt
OCRExtract text from images or documentsRead license plates, receipts, screenshots
Face detectionDetect face locations and visible attributesCount faces in an image
Face analysisAnalyze face-related attributes where permittedEstimate visible facial characteristics

Image Classification ​

Image classification answers "what is this image?" It can be:

  • Multiclass: one label per image.
  • Multilabel: multiple labels per image.

Object Detection ​

Object detection answers "what objects are present and where are they?" It returns labels plus bounding boxes.

Choose object detection when the scenario mentions:

  • Counting items in an image.
  • Locating each item.
  • Drawing boxes around detected items.

Optical Character Recognition ​

OCR extracts printed or handwritten text from images and scanned files. Use OCR for screenshots, scanned documents, signs, labels, forms, and receipts when the task is to read text.

Face Detection and Analysis ​

Face detection locates faces in an image. Face-related capabilities are sensitive and may require approval, responsible use, and compliance with Microsoft's policies and local law.

Azure Tools and Services ​

Azure AI Vision ​

Azure AI Vision provides image analysis capabilities such as:

  • Image tagging and categorization.
  • Caption and dense caption generation.
  • Object detection.
  • OCR/read text extraction.
  • Smart cropping and visual feature extraction.

Use Azure AI Vision for general image analysis and OCR.

Azure AI Face ​

Azure AI Face provides face detection and face-related capabilities. For exam purposes, distinguish it from general image analysis:

  • Azure AI Vision: general image understanding and OCR.
  • Azure AI Face: face-specific detection and analysis features.

Decision Rules ​

text
Label entire image                 -> Image classification
Find and locate multiple objects   -> Object detection
Extract text from image/document   -> OCR
Detect or analyze faces            -> Azure AI Face
General image analysis             -> Azure AI Vision

Exam Traps ​

  • OCR vs document processing: OCR reads text; document processing extracts structured fields from forms and documents.
  • Classification vs object detection: Classification labels; object detection locates.
  • Vision vs Face: General image features map to Azure AI Vision; face-specific scenarios map to Azure AI Face.
  • Responsible use matters: Face capabilities are sensitive; do not ignore consent, policy, and compliance requirements.

Flashcards ​

Flashcards

1 / 4
❓

Which vision workload returns bounding boxes?

(Click to reveal)
💡
Object detection.

<- Previous Domain | Overview | Next Domain ->