Teachable Machine
What you'll learn
- Train your own image classification model without writing code
- Understand how machine learning models learn from examples
- See firsthand why the quality of training data matters
What is Teachable Machine?
Teachable Machine is a free tool from Google that lets you train your own machine learning model without writing any code. You'll teach a computer to recognise different things by showing it examples.
This is the same basic principle used in more complex AI systems – they all learn by seeing lots of examples!
Ready to Start?
Click the button below to open Teachable Machine in a new tab. Keep this page open to track your progress.
Step-by-Step Guide
Create Your Classes
Think of two or more things you want the AI to recognise. For example: "thumbs up" and "thumbs down", or "happy face" and "surprised face".
Add Training Samples
Use your webcam to record examples of each class. Aim for at least 20-30 samples per class. Move around, change angles, and vary the lighting.
Train Your Model
Click "Train Model" and wait. The computer is learning patterns from your examples. This usually takes less than a minute.
Test It Out
Show the camera new examples and see if it classifies them correctly. Try things it hasn't seen before!
Experiment and Improve
Try to confuse your model. What happens if you show it something completely different? Add more samples to improve accuracy.
⚠️ Common Mistakes
- •Too few samples: The model needs enough examples to learn patterns. Aim for 20+ per class.
- •No variety: If all your samples look the same (same angle, same lighting), the model won't generalise well.
- •Background matters: The model might learn to recognise your background instead of the object!
- •Similar classes: Very similar categories (like two similar hand positions) are harder to distinguish.