1. Computer Vision Models (Analysis of Images & Video)
✅ What they do
Computer vision models interpret visual data (images/video) and extract meaning.
🔑 Key capabilities (you must recognise in the exam)
1. Image Classification
Assigns a label to an entire image
Example: “This is a dog”
👉 Use when:
One main subject per image
High-level categorisation
2. Object Detection
Identifies multiple objects + their locations (bounding boxes)
Example: Detect cars, people, traffic lights in one image
👉 Use when:
You need position + type of objects
3. Image Segmentation
Classifies each pixel
More precise than object detection
👉 Use when:
Fine-grained analysis (e.g. medical imaging)
4. Facial Analysis (not identification in AI-901 context)
Detects:
Faces
Emotions
Landmarks (eyes, nose, etc.)
👉 Important exam nuance:
Face detection ≠ face recognition
5. Optical Character Recognition (OCR)
Extracts text from images
👉 Use when:
Scanned documents
Receipts, invoices, forms
6. Image Tagging & Captioning
Generates descriptions or tags
Example: “A person riding a bike”
🧠 Model selection logic (AI-901 mindset)
Scenario Correct Model
“What is in this image?” Classification
“Where are the objects?” Object Detection
“Read text from image” OCR
“Describe the image” Captioning
🎨 2. Image Generation Models (Generative AI)
✅ What they do
These models create new images, rather than analysing existing ones.
🔑 Key capabilities
1. Text-to-Image Generation
Input: Text prompt
Output: New image
Example:
“A futuristic London skyline at sunset”
👉 Used in:
Marketing
Design
Prototyping
2. Image Editing / Inpainting
Modify parts of an image
Fill missing areas intelligently
3. Style Transfer
Apply artistic styles (e.g. Van Gogh effect)
4. Image Variation
Generate multiple versions of an image
5. Synthetic Image Creation
Generate training data (e.g. fake faces, objects)
🧠 Model selection logic
Scenario Correct Model
“Create an image from text” Text-to-image model
“Edit part of an image” Inpainting
“Apply artistic style” Style transfer
“Generate training images” Generative model
⚖️ Key Differences (Exam Gold)
Feature Computer Vision Image Generation
Purpose Analyse images Create images
Input Image/video Text or image
Output Labels, objects, text New images
Example OCR, detection DALL·E-style models
🎯 AI-901 Exam Tip (Important)
You’ll often be given a scenario, like:
“A system needs to extract text from scanned invoices”
✔ Correct answer: OCR (Computer Vision)
❌ Not: Generative AI
🚀 Simple Memory Trick
Vision = Understand images
Generation = Create images