AI Vision with PictoBlox, lesson 15 of 15
Final project: build your own AI helper
Introduction
You can now make a computer see motion, faces, expressions, bodies, hands, objects and words. You've trained your own model and made AI switch on a real light. Time to put it all together and invent something of your own: an AI helper that solves a real problem or makes people smile.
Your recipe: when the camera sees a hand sign, the hand pose model decides, and the helper Arduino turns on a light. Every AI project is these three steps: sense, think, act.
Working principle
Every great project follows the same steps, the way real engineers work:
- Problem: what will your helper do, and who is it for?
- Plan: fill in the see → think → act steps.
- Build: start with the smallest version that works, then add to it.
- Test: try it with different people and conditions (remember the fairness test!).
- Improve: fix what didn't work.
- Show: present it to your family or class.
What you'll need
- PictoBlox with whichever extensions your idea needs
- A webcam
- Optional: an Arduino, LEDs, a buzzer or a servo motor
- A notebook for your plan
Working in real life
Step 1: pick an idea. Use one of these, or invent your own:
| Idea | See | Think | Act |
|---|---|---|---|
| Study buddy | Your face | Are you looking at the screen, or away for too long? | Tobi reminds you to focus |
| Clean-up game | Objects on your desk | Is there a cup, bottle or book left out? | A score goes up for each thing you tidy away |
| Exercise coach | Your body pose | Did your hands go above your head? | Counts your jumping jacks out loud |
| Mood lamp | Your face | Happy, sad or surprised? | An Arduino LED changes colour |
| Sign-language helper | Your hand | Which sign is it? (train your own model) | Tobi says the word |
| Recycling sorter | An item held up to the camera | Paper, plastic or metal? (train your own model) | Tells you which bin to use |
Step 2: write your plan. On one page, write: the name of your helper, who it helps, its see → think → act steps, which extensions you'll use, and how you will test it.
Step 3: build the smallest version. Get just the "see" and "think" part working first. Make Tobi say what the AI sees. Then add the "act" part.
Step 4: test and improve. Ask three people to try it. Write down what went wrong and fix one thing at a time.
Step 5: show it off. Give a two-minute demo: what problem it solves, a live demo, and one thing you'd improve next.
Try it
When your project works, add one extra challenge:
- Add sound effects and a title screen.
- Add a real-world part with an Arduino.
- Make it work for two people at once.
- Make it fairer: test it in a dim room and with someone wearing glasses, and fix what fails.
Check what you learnt
What should you build first?
Your exercise coach miscounts when your friend tries it. What's the best next step?
Which part of 'see, think, act' is an LED changing colour?
Conclusion
- Real projects follow the same steps: problem, plan, build, test, improve, show.
- Start small, get it working, then add more.
- You're now an AI maker: you can make computers see, think and act!
What next?
Want to go further? When you're older, try the AI & Computer Vision course, where you'll build the same kinds of projects in Python with OpenCV and YOLO. Until then, keep inventing!

