How does a machine see?

Learning goal

Explain why varied examples and careful testing matter in computer vision.

What you’ll need

• Paper and drawing materials
• A selection of everyday object pictures
• Optional access to a classroom image-classification tool

The activity

1. Choose two object categories. Draw or collect several different examples of each.
2. Write a rule that separates the categories. Try the rule on a new picture.
3. Find a confusing example and explain why it challenges the rule.
4. If using a classification tool, test new examples with different angles or backgrounds and record mistakes.

Pause and reflect

• Which examples were easy to recognise?
• How did the background or angle affect the result?
• When should a person check a machine’s decision?

Take it further

Add a third category and compare how the classification challenge changes.