Level 1 • Easy
🕵️ Activity 1: Be a Human CNN
Your brain already works a bit like a CNN! Look at each picture and type the clues (features) that help you recognize it.
Level 2 • Easy
🧱 Activity 2: Explore CNN Layers
A CNN looks at an image in steps (layers). Click each layer to see what it does!
🔍 Layer 1 – Edges & Lines
🔷 Layer 2 – Simple Shapes
🧩 Layer 3 – Object Parts
🏷 Final Layer – Prediction
Click any layer above to learn what it does!
Level 3 • Medium
📸 Activity 3: Explore the Dataset
AI learns from many example pictures. Look at the four groups and answer the questions.
Level 4 • Medium
🛠️ Activity 4: Build Your CNN
Click the pieces in the correct order to assemble a simple CNN. (Think of the journey of an image!)
Pieces placed: 0 / 5
📸 Image Input
🔍 Convolution (find features)
🧊 Pooling (keep important bits)
📄 Flatten
🏷 Prediction
Level 5 • Challenge
📊 Activity 5: AI Doctor – Improve Accuracy
Your AI got only 60% accuracy. Choose the best ways to make it better!
Current Accuracy: 60%
📷 More Images
✨ Better Quality
🔄 Different Angles
⏱ Train Longer
⚖️ Balance Dataset
🗑 Remove Bad Labels
Click up to 3 improvement cards, then press Boost!
Level 6 • Design
🚀 Activity 6: Design Your Own Image AI
Invent a useful image-recognition AI for school, home or community.
🏅
Project Showcase
Professor Spark awards you the CNN Explorer & AI Visionary badges!
Level 7 • Advanced
🧬 Activity 7: Meet the CNN Families
Not all CNNs are the same! Click each family to discover what makes it special.
🧱
Classic CNN
Simple stack of layers (like the one you built earlier)
🗼
Deep CNN (VGG-style)
Many layers stacked – more power, more work
🔗
Residual (ResNet)
Uses “skip connections” so information can jump
📱
Lightweight (MobileNet)
Tiny & fast – perfect for phones (TensorFlow Lite!)
Click a CNN family above to learn more!
Level 8 • Advanced
🔗 Activity 8: Skip Connection Challenge
In ResNet, some information can skip layers so the network doesn’t forget early details. Choose which path helps the signal survive!
A deep stack of layers. Click the blocks that should get a skip connection (green path).
Layer 1
→
Layer 2
→
Layer 3
→
Layer 4
→
Layer 5
Hint: Skipping over the middle layers often helps the deepest networks.
Level 9 • Expert
🎯 Activity 9: Match the Task to the Best CNN
Different jobs need different CNN styles. Choose the best fit for each mission.
📷 “Is this a cat or a dog?” (simple classification)
🏆 Winning a hard image contest (needs maximum accuracy)
📱 App on a phone that recognizes plants instantly
🔬 Very detailed features (many stacked filters)