Build & Test a Model

Your mission

Build a real (small) machine-learning model with a free browser tool, then test it honestly on data it never trained on.

🎒What you'll need

• ▸A computer with internet
• ▸A free tool such as Google's Teachable Machine
• ▸About 30 example images or sounds you gather

🧭Your step-by-step plan

1. 1

   Choose a task

   Decide what your model recognises — thumbs-up versus thumbs-down, or cat versus dog drawings.

2. 2

   Gather data

   Collect at least 15 examples per class, keeping a few aside as a test set the model never trains on.

3. 3

   Train the model

   Feed your training examples into the tool and let it train.

4. 4

   Test and measure

   Run your held-out test examples through it and write down the accuracy.

✨Make it yours

• ★Add a third class and see how accuracy changes.
• ★Train it on deliberately messy data and explain what happens.

✅How you'll know you succeeded

• ✓The model trained on data you collected yourself.
• ✓Accuracy was measured on a separate test set.
• ✓You can explain one way to make it more accurate.