Curriculum Overview: Identifying Computer Vision Workloads

This document outlines the training path for mastering computer vision workloads as defined by the Microsoft Azure AI Fundamentals (AI-900) curriculum. Computer vision is a critical domain of Artificial Intelligence that enables software to interpret and understand visual information from the world, representing approximately 15–20% of the AI-900 exam content.

Prerequisites

Before diving into computer vision workloads, learners should possess a foundational understanding of the following:

• Basic Cloud Concepts: Familiarity with cloud computing models (IaaS, PaaS, SaaS) and general Azure architecture.
• General AI Awareness: Understanding that AI models are trained using data and used to make predictions.
• Data Literacy: Basic knowledge of digital file types (JPEG, PNG, MP4) and how data is structured.
• Mathematics Basics: High-level awareness of coordinate systems ($x, y$ coordinates) which are used for bounding boxes in image analysis.

Module Breakdown

The curriculum is structured into a logical progression, starting from high-level workload identification to specific Azure service implementation.

Learning Objectives per Module

Module 1: Foundations of Computer Vision

• Define the concept of "Machine Seeing."
• Understand how digital images are processed as arrays of pixel values.

Module 2: Image Analysis and Classification

• Image Classification: Identifying the primary subject of an image (e.g., "This is a car").
• Image Tagging: Generating descriptive metadata for an image.

Module 3: Object Detection and Spatial Analysis

• Object Detection: Identifying individual objects within an image and providing their location via bounding boxes.
• Semantic Segmentation: Mapping specific pixels to specific objects (more granular than object detection).

Module 4: Optical Character Recognition (OCR)

• Use Azure AI Vision to extract printed or handwritten text.
• Differentiate between simple OCR and Document Intelligence.

Module 5: Facial Detection and Analysis

• Distinguish between Facial Detection (finding a face) and Facial Recognition (identifying who the face belongs to).
• Analyze facial attributes (age, emotion, glasses).

Success Metrics

To demonstrate mastery of this curriculum, the learner must be able to:

1. Workload Identification: Correctly match a business scenario (e.g., "A retailer needs to count people entering a store") to the appropriate AI workload (Object Detection).
2. Service Selection: Choose between Azure AI Vision and Azure AI Face based on the specific requirements of the project.
3. Concept Differentiation: Explain the difference between classification (what) and object detection (what and where).
4. Responsible AI Application: Identify potential biases in facial analysis and suggest mitigation strategies according to Microsoft's guiding principles.

[!IMPORTANT] Success in the AI-900 exam requires knowing that Azure AI Vision is the primary multi-purpose service, while Azure AI Face is a specialized service for facial analysis.

Real-World Application

Computer vision is no longer theoretical; it is embedded in modern industry. Understanding these workloads allows for the following implementations:

• Manufacturing: Using Object Detection on assembly lines to identify defective parts in real-time.
• Healthcare: Applying Image Classification to X-rays or MRI scans to assist radiologists in identifying anomalies.
• Retail: Implementing OCR to automatically scan receipts for loyalty point programs.
• Public Safety: Utilizing Facial Detection in smart doorbells to alert homeowners when a person (rather than a pet) is at the door.

[!TIP] When studying for the exam, focus on the outputs of each workload. If the output includes a set of coordinates, it is almost certainly Object Detection or Image Segmentation.