Equipment Needed
- Windows PC with Chrome or Edge browser
- Microsoft Paint
- A folder on the desktop called TM_Lesson5
Key Vocabulary
Starter Activity (5 minutes)
Brainstorm: Think of 3 real-world problems that could be solved by a computer that can recognise images. Write them below.
Example ideas: Sorting recycling (plastic vs paper vs glass), identifying dog breeds, spotting weeds in a garden, reading handwritten numbers, sign language recognition, checking if PPE is being worn.
Main Tasks
Part A: Plan Your Product
Choose one idea and plan your product. Fill in the design sheet below:
My AI Product Design Sheet
Support: Students who struggle to choose can use one of the brainstorm ideas from the starter. Have a few pre-made suggestions on the board. Encourage simple, achievable ideas (2–3 classes max).
Part B: Build Your Prototype
-
Create your training images in Paint. Draw at least 10 images per class. Save them in organised folders inside
TM_Lesson5. - Open Teachable Machine and create a new Image Project.
- Set up your classes and upload your training images.
- Train the model.
- Create separate test images (at least 3 per class) and test your model.
Part C: Record & Evaluate
| # | Test Image | Actual Class | Prediction | Correct? |
|---|---|---|---|---|
| 1 | ||||
| 2 | ||||
| 3 | ||||
| 4 | ||||
| 5 | ||||
| 6 |
Evaluation
Common Mistakes & Quick Fixes
- Idea too complicated: Keep it simple — 2 or 3 classes is enough. Don't try to build something with 10 classes.
- All images look the same: Vary size, position, and thickness of your drawings. This helps the model generalise.
- Running out of time: Focus on getting the model working first, then do the evaluation. You can improve the model for homework or next lesson.
Challenge / Stretch Task
Create a poster (in Paint or on paper) advertising your AI product. Include: product name, what it does, who it's for, and how accurate it is.
Exit Ticket
Assessment focus: Look for students who can explain their product clearly, give a realistic accuracy figure, and suggest meaningful improvements (not just "make it better").