Lesson 7: How AI Sees Images

PRINTING INSTRUCTIONS

Materials Needed:

• Cardstock paper (recommended for durability)
• Color printer
• Scissors or paper cutter
• Optional: Laminating sheets for extended use

Print Settings:

• Paper size: 8.5" × 11" letter
• Print quality: High/Best
• Color: Full color
• Double-sided: No
• Pages needed: 10 pages (3 cards per page = 30 cards total)

Assembly:

1. Print all pages on cardstock
2. Cut along dotted lines to separate cards
3. Optional: Laminate cards for durability
4. Store in envelope or zip-lock bag labeled "Lesson 7 Image Cards"

Card Size: Each card measures approximately 3.5" × 4.5" when cut

HOW TO USE THESE CARDS

Activity Purpose: Students act as "rule-based AI" and sort images according to specific classification rules you provide.

Suggested Rules for Sorting:

Rule Set 1 - Color Based:

• Group A: Mostly warm colors (red, orange, yellow)
• Group B: Mostly cool colors (blue, green, purple)
• Group C: Mostly neutral (black, white, gray, brown)

Rule Set 2 - Shape Based:

• Group A: Contains mostly circles or curves
• Group B: Contains mostly straight lines and angles
• Group C: Mix of both

Rule Set 3 - Content Based:

• Group A: Living things (plants, animals, people)
• Group B: Human-made objects (buildings, vehicles, tools)
• Group C: Natural non-living (landscapes, sky, water)

Rule Set 4 - Complexity:

• Group A: Simple (1-2 main objects, plain background)
• Group B: Moderate (3-5 objects or elements)
• Group C: Complex (many objects, busy scene)

CARD DESCRIPTIONS

PAGE 1: Animal Images

Card 1: GOLDEN RETRIEVER

• Image: Happy golden retriever sitting, facing camera
• Colors: Golden/tan, white background
• Features: Clear dog features, four legs visible, friendly expression
• Intentional Use: Easy to classify, standard dog image

Card 2: CHIHUAHUA CLOSE-UP

• Image: Close-up of chihuahua face, big eyes, tan/brown coloring
• Colors: Brown, tan, black
• Features: Unusual angle, very close, emphasizes large eyes and small snout
• Intentional Use: Different from typical dog images, tests breed recognition

Card 3: BLACK CAT SILHOUETTE

• Image: Black cat sitting, backlit so it appears as silhouette
• Colors: Black against light background
• Features: Cat shape recognizable but no detail, completely dark
• Intentional Use: Tests recognition without color or texture detail

PAGE 2: Food Images (Trick Cards)

Card 4: BLUEBERRY MUFFIN TOP

• Image: Close-up of muffin top, textured, rounded, brown with blue spots
• Colors: Brown, tan, blue
• Features: Round, textured surface, blueberries look like eyes/nose
• Intentional Use: Famous AI confusion—can look like a dog face!

Card 5: BAGEL WITH SEEDS

• Image: Round bagel photographed from angle, sesame seeds on top
• Colors: Tan, brown, white
• Features: Round, textured, holes in middle
• Intentional Use: Can be confused with dog snout at certain angles

Card 6: CHOCOLATE CHIP COOKIE

• Image: Round cookie, chocolate chips visible
• Colors: Brown, tan, dark brown chips
• Features: Round shape, spotted pattern
• Intentional Use: Tests pattern recognition—spots could confuse AI

PAGE 3: Nature - Warm Colors

Card 7: SUNSET OVER OCEAN

• Image: Orange and red sunset with water reflecting colors
• Colors: Orange, red, yellow, some blue
• Features: Horizontal layers of color, natural scene
• Intentional Use: Primarily warm colors but has cool elements

Card 8: AUTUMN LEAVES

• Image: Close-up of red and orange maple leaves
• Colors: Red, orange, yellow, brown
• Features: Organic shapes, overlapping, nature pattern
• Intentional Use: Clear warm color dominant

Card 9: RED ROSE

• Image: Single red rose bloom, close-up
• Colors: Red, green stem, green leaves
• Features: Circular flower shape, organic, living thing
• Intentional Use: Living thing with primarily warm color

PAGE 4: Nature - Cool Colors

Card 10: BLUE SKY WITH CLOUDS

• Image: Clear blue sky with white puffy clouds
• Colors: Blue, white
• Features: Amorphous shapes, gradients, natural
• Intentional Use: Clear cool colors, but minimal detail

Card 11: GREEN FOREST

• Image: Dense forest scene with green trees
• Colors: Various shades of green, brown trunks
• Features: Complex, many vertical lines, organic
• Intentional Use: Cool color dominant, complex scene

Card 12: PURPLE FLOWERS

• Image: Cluster of purple lavender flowers
• Colors: Purple, green stems
• Features: Repeated small shapes, organic pattern
• Intentional Use: Living thing with cool colors

PAGE 5: Geometric Objects

Card 13: RED STOP SIGN

• Image: Standard octagonal stop sign, clear view
• Colors: Red, white letters
• Features: Eight-sided, geometric, clear text, human-made
• Intentional Use: Standard shape AI should recognize

Card 14: YELLOW CIRCULAR WARNING SIGN

• Image: Round yellow traffic sign
• Colors: Yellow, black symbol
• Features: Perfect circle, geometric, high contrast
• Intentional Use: Simple geometric shape

Card 15: BASKETBALL

• Image: Orange basketball against white background
• Colors: Orange, black lines
• Features: Sphere, curved lines creating pattern
• Intentional Use: Circular object with distinctive pattern

PAGE 6: Buildings and Architecture

Card 16: MODERN GLASS BUILDING

• Image: Contemporary skyscraper with glass facade
• Colors: Blue (reflected sky), gray, silver
• Features: Vertical lines, geometric, reflective, human-made
• Intentional Use: Cool colors, geometric, modern

Card 17: RED BRICK BUILDING

• Image: Traditional brick building facade
• Colors: Red, brown, white windows
• Features: Rectangular pattern, geometric, traditional
• Intentional Use: Warm colors, geometric pattern

Card 18: WHITE HOUSE

• Image: Small white house with pitched roof
• Colors: White, some blue (sky), green (grass)
• Features: Simple structure, clear building, basic shapes
• Intentional Use: Mixed colors, clear human-made structure

PAGE 7: Black and White Images

Card 19: ZEBRA

• Image: Zebra showing distinctive stripes
• Colors: Black and white stripes
• Features: Living thing, high contrast, strong pattern
• Intentional Use: No color to classify by—tests other features

Card 20: DALMATIAN DOG

• Image: White dog with black spots
• Colors: White, black spots
• Features: Dog shape, spotted pattern
• Intentional Use: Black and white but is a dog—interesting case

Card 21: NEWSPAPER PAGE

• Image: Black and white newspaper with text and photos
• Colors: Black text, white paper, gray photos
• Features: Rectangular, text, human-made
• Intentional Use: No color, complex detail, human-made object

PAGE 8: Complex Scenes

Card 22: BEACH SCENE

• Image: Beach with people, umbrellas, ocean, sky
• Colors: Blue (sky/water), tan (sand), various (people/objects)
• Features: Multiple objects, complex, both natural and human-made
• Intentional Use: Cannot be sorted into single simple category

Card 23: CITY STREET

• Image: Busy street with cars, buildings, people, signs
• Colors: Mixed—all colors present
• Features: Very complex, many objects, all human-made
• Intentional Use: Overwhelming detail, tests complexity sorting

Card 24: GARDEN

• Image: Flower garden with many different colored flowers
• Colors: All colors—red, yellow, purple, pink, orange
• Features: Organic, living things, complex patterns
• Intentional Use: Living things but all colors present

PAGE 9: Ambiguous/Tricky Images

Card 25: PERSON LYING DOWN FROM ABOVE

• Image: Person lying on ground photographed from directly overhead
• Colors: Various (clothing), skin tone
• Features: Unusual perspective, person but not typical view
• Intentional Use: Living thing but unusual angle—could be misclassified

Card 26: PARTIALLY HIDDEN DOG

• Image: Dog mostly behind tree, only tail and partial body visible
• Colors: Brown (dog), green (grass), brown (tree)
• Features: Incomplete object, context needed
• Intentional Use: Tests what happens when only part of object visible

Card 27: EXTREME CLOSE-UP OF ORANGE

• Image: Very close photo of orange peel texture
• Colors: Orange
• Features: Texture only, no shape context, organic
• Intentional Use: Too close to see what object is—tests need for context

PAGE 10: Additional Categories

Card 28: BLUE CAR

• Image: Modern blue sedan, side view
• Colors: Blue, black (tires), chrome
• Features: Geometric, human-made, vehicle, clear shape
• Intentional Use: Clear cool color, human-made object

Card 29: YELLOW SCHOOL BUS

• Image: Classic yellow school bus
• Colors: Yellow, black trim, white
• Features: Geometric, human-made, vehicle, distinctive
• Intentional Use: Clear warm color, iconic vehicle

Card 30: HANDS HOLDING SMARTPHONE

• Image: Close-up of hands holding phone, screen visible but generic
• Colors: Skin tone, black (phone), various (screen)
• Features: Human element, technology, some organic and some geometric
• Intentional Use: Combines living thing and object, modern context

TEACHER GUIDE FOR USING THE CARDS

Setup (3 minutes)

1. Divide class into groups of 3-4 students
2. Give each group one complete set of shuffled cards (face down)
3. Write classification rules clearly on board
4. Explain: Students must sort cards using ONLY the rules, not their human judgment

Activity Flow (8-12 minutes)

Round 1: Basic Sorting (4 minutes)

• Groups flip cards one at a time
• Classify each card into Group A, B, or C based solely on stated rules
• If ambiguous, place in "Not Sure" pile
• No discussion of what image "really" shows—only which rule it matches

Round 2: Reveal and Discuss (4 minutes)

• Groups examine their sorted piles
• Discuss surprising classifications
• Identify "trick cards" that don't fit neatly
• Share examples with class

Round 3: Reflection (3-4 minutes)

• Connect to AI limitations: "This is how rule-based AI works!"
• Discuss: What was frustrating? What would help AI do better?
• Compare to Teachable Machine (pattern learning vs. rigid rules)

Discussion Questions

For Teachers to Ask:

1. "Which cards were hardest to classify? Why?"
2. "Did any cards belong in multiple groups? What does this teach us about AI?"
3. "How would a human classify these differently? What can humans consider that AI cannot?"
4. "What happened with black-and-white images when rules focused on color?"
5. "How do the 'trick cards' (muffin, bagel) relate to real AI mistakes?"

Common Student Observations

Students often notice:

• Rules don't work for ambiguous cases (beach scene has all colors)
• Partial images are impossible to classify (hidden dog)
• Black-and-white images break color-based rules
• Similar patterns can fool the classification (muffin looks like dog)
• Human understanding of context would help but rules don't allow it

Extensions

Create Your Own Rules:

Let student groups create their own classification rules and swap with another group to test them.

Add More Cards:

Have students bring their own images or draw cards that would be "easy" or "hard" for AI to classify.

Multiple Rule Sets:

Sort the same cards using different rules and compare results—shows AI behavior depends on what rules it's given.

Cards Requiring Special Discussion

Muffin (Card 4) and Bagel (Card 5):

Explain this is a real AI problem—show actual examples of AI confusing these with dogs!

Black and White Cards (19-21):

If using color-based rules, these become impossible—great teaching moment about AI limitations.

Complex Scenes (22-24):

These don't fit neatly anywhere—perfect for discussing why context matters.

Partial/Unusual Views (25-27):

Demonstrate that AI needs to see enough of an object and from normal perspectives.

ALTERNATIVE PRINTING OPTIONS

Option 1: Full Color on Cardstock (Recommended)

• Best quality and durability
• Cards stand up to repeated use
• Colors are vivid for clear classification

Option 2: Grayscale/Black and White

• More affordable for large classes
• Works well for shape-based or content-based rules
• Cannot use color-based rules

Option 3: Digital Display

• Project cards on screen one at a time
• Students classify as whole class or small groups
• Saves printing but less tactile engagement

Option 4: Student-Created Cards

• Have students find or draw their own images
• Mount on index cards
• Takes more time but increases engagement and ownership

STORAGE AND REUSE

Organizing Sets:

• Each complete set should be stored together
• Use envelopes, zip-lock bags, or small boxes
• Label: "Lesson 7 - Image Classification Cards - Set 1, Set 2," etc.
• You'll need one set per group of 3-4 students

Durability Tips:

• Laminate with 3-5 mil laminating pouches
• Round corners after cutting to prevent peeling
• Store flat to prevent bending
• Keep out of direct sunlight to prevent fading

Reuse Suggestions:

• Use in other AI lessons about pattern recognition
• Use for review activities or warm-ups
• Let students create their own sorting challenges
• Adapt for assessment activities

ACCESSIBILITY MODIFICATIONS

For Students with Visual Impairments:

• Create tactile cards with raised features
• Provide verbal descriptions of each card
• Use cards with very high contrast
• Partner with sighted peer

For Students with Motor Challenges:

• Reduce number of cards (use 15 instead of 30)
• Make cards larger (5" × 7")
• Allow verbal sorting instead of physical manipulation
• Provide card holder or stand

For English Language Learners:

• Use cards with clear, simple images
• Allow discussion in native language within groups
• Focus on visual features rather than verbal descriptions
• Pair with bilingual buddy

ASSESSMENT OPPORTUNITIES

Observe for Understanding:

• Are students following rules strictly or using human judgment?
• Can they articulate why certain cards are ambiguous?
• Do they connect activity to AI limitations?
• Are they using vocabulary correctly (pattern, feature, classification)?

Exit Ticket Questions:

• "Which card was hardest to classify and why?"
• "How does this activity show the difference between AI and human thinking?"
• "What would help an AI classify these images better?"

Evolve AI Institute • Lesson 7: How AI Sees Images

Printable Image Cards - Teacher Guide

Note to Teachers: While we describe the content of cards in detail, you'll need to source or create actual images matching these descriptions. Copyright-free image sources include:

• Unsplash.com (free high-quality photos)
• Pixabay.com (free images and vectors)
• Pexels.com (free stock photos)
• Public domain collections (NASA, government images)
• Creative Commons licensed images (with attribution)

Or use AI image generators with appropriate filters, hand-drawn cards by students, or cut images from magazines with proper permissions.