← All Sessions

Image Recognition
Through Building

AI Trade School - Session 3A

How Machines See & Classify Images

🔄 Welcome Back

You've Built AI 4 Times Already

✍️

Week 1 Prompts

AI fundamentals mastered

📁

Week 2 Models

ML principles understood

🧠

4 Portfolio Pieces

Professional credentials building

🎤

Week 3A Start

Computer Vision Specialist

Why Do Some AI Systems
See Better Than Others?

Computer Vision
is Pattern Recognition

Just With Images Instead of Numbers

🎯 Computer Vision In Action

🏥

Medical Imaging

Detect tumors in X-rays and scans

🏭

Manufacturing

Spot defects on assembly lines

🔒

Security

Face recognition and monitoring

🌾

Agriculture

Detect crop disease from drones

Every single one solves business problems worth billions.

5 CAREER PATHWAYS

Real Jobs in South Carolina

Healthcare AI

Train models to recognize diseases

$65K-$95K

Manufacturing

Spot defects on assembly lines

$55K-$85K

Retail Analytics

Understand customer behavior

$50K-$75K

Ag-Tech

Recognize crop health from drones

$55K-$85K

Security

Face recognition and surveillance

$60K-$95K

More Opportunities

Autonomous systems, robotics, analytics

Growing Field

How Machines Actually
See Images

YOU SEE: A complete image instantly

MACHINE SEES: Thousands of pixel values (0-255)

TOGETHER: Human creativity + Machine consistency = Powerful

🧠 The ML Learning Process

Four Steps to Pattern Recognition

1
SEE EXAMPLES
1,000+ labeled images
2
FIND PATTERNS
What distinguishes each class
3
LEARN FEATURES
Adjust weights precisely
4
PREDICT NEW
Classify unseen images

Computer Vision Solves
Real Business Problems

Hospital: 8 hours → 2 hours image review. That's 6 hours saved daily.

Factory: One recall prevented = millions saved. AI checks 100 items/min.

Retailer: Customer behavior insights = smarter layouts = more sales.

Three Phases. Three Hours.

1
TEACH
50 Minutes
2
BUILD
110 Minutes
3
SHOWCASE
20 Minutes

🏆 Portfolio Piece #5

Trained Image Classification Model

What It Is

Real ML model trained on your dataset

What It Does

Classifies images with documented accuracy

Why It Matters

Employers recognize this immediately

Your Value

Worth $60K-$80K starting salary

YOU CAN
DO THIS

Proof: You've Done It Before

Week 1 felt impossible. You figured it out.
Week 2 was too technical. You nailed it.
Week 3? Same pattern. Same outcome.

YOU'VE GOT THIS.

NOW WE BUILD

110 Minutes. Your Image Model.

Step 1: Setup

Choose use case, load dataset (15 min)

Step 2: Configure

Tell platform what to build (20 min)

Step 3: Train

Model learns patterns (40 min)

Step 4: Evaluate

Check accuracy, document (25 min)

💡 Choose Your Problem

Multiple Real-World Options

Medical: Classify X-rays (pneumonia vs. normal)

Manufacturing: Detect defects (okay vs. defective)

Documents: Classify type (invoice, receipt, contract)

Agriculture: Plant health (healthy, diseased, stressed)

Retail: Product condition (new vs. used)

Security: Detect objects (person, vehicle, animal)

Wildlife: Classify animals or land use

📊 Step 1: Prepare Dataset

15 Minutes

UPLOAD

Place training images in folders

ORGANIZE

Training, validation, test split

VERIFY

Images must be .jpg or .png

CONFIRM

You have 3+ classes defined

⚙️ Step 2: Configure Model

20 Minutes

Name: FirstName_LastName_ImageClassification

Specify classes: "Normal, Pneumonia, Other"

Training budget: 0.5-1.0 hours

Enable augmentation to prevent overfitting

Click START TRAINING

Model Training

What's Happening (40 Minutes)

Step 1: Analyze 1,000+ training images

Step 2: Find patterns that separate classes

Step 3: Adjust weights until highly accurate

Step 4: Validate learning on test data

🧩 Extension Activities

While Model Trains (40 Minutes)

Activity 1: What mistakes might your model make? Why?

Activity 2: What features matter to your model?

Activity 3: How do you ensure fairness across demographics?

These aren't homework. They're real ML challenges professionals face.

📈 Step 3: Evaluate Results

25 Minutes

ACCURACY

% classified correctly (70%+ is good)

PRECISION

When it predicted X, was it right?

RECALL

How many real cases did it catch?

CONFUSION

What did it get confused about?

📝 Step 4: Document Results

10 Minutes

Problem statement: What are you solving?

Dataset details: Images, classes, split

Model performance: Accuracy, precision, recall

What worked well: Specific successes

Business application: How would companies use this?

✨ Polish & Showcase Prep

15 Minutes

Screenshot

Accuracy metrics and confusion matrix

Organize

All documentation filled in completely

Practice

Say your pitch out loud to pod

Refine

Get peer feedback and improve

SHOWCASE TIME

Celebrate Your Work

3-4 of you will share what you built

2 minutes each

Everyone listens and gives feedback

Peer validation is powerful. When your classmate says "that's cool," it means something.

🎯 What You Built Today

Recognition of Achievement

Trained a real image classification model
Tested it on images it's never seen
Evaluated performance with real metrics
Documented results professionally
Understood why it succeeded or struggled

THAT'S SPECIALIST-LEVEL AI SKILL

You're Not Learning
ABOUT Computer Vision

YOU ARE
COMPUTER VISION PEOPLE

5 portfolio artifacts built
You understand WHEN to use AI
You know HOW to build it
You can evaluate WHETHER it works

📚 Next Week: We Go Deeper

Build Your First Generative Model

How machines CREATE new content

Keep Your Documentation
Show It In Every Interview

Real Data Is
Always Complex

You Just Learned How to Understand It

That's a Professional Skill

Questions?
1 / 25
Title
Welcome Back
Why CV
Applications
Careers
How See
ML Process
Business Impact
Today's Plan
Artifact
Confidence
Build Intro
Use Cases
Dataset Prep
Configure
Training
Activities
Evaluate
Document
Polish
Showcase
Built
Superpower
Forward
Final
/ Space next
back
F fullscreen