Build a Teachable Machine Image Classifier With Your Kid

Issue 43 • August 14, 2024

Introducing children to artificial intelligence can feel daunting, especially when trying to move them from passive consumers of technology to active creators. Many educational tools require complex coding skills or expensive software subscriptions.

Google's Teachable Machine offers a free, web-based alternative where kids can train a real machine learning model in minutes using a standard computer webcam. This hands-on project demystifies how computers learn to recognize objects, gestures, and patterns, providing a perfect weekend activity for busy Executives and their children.

The Project: Build a Custom Object Detector

In this activity, your child will build an image classifier that can distinguish between two different items or hand gestures. For example, they can train the computer to recognize a coffee mug versus a water bottle, or a thumbs-up versus a thumbs-down gesture.

This project requires no coding, no account creation, and no specialized hardware. All you need is a computer with a working webcam and an internet connection. By actively capturing the training data and testing the results, your child will gain a practical understanding of supervised learning.

Step 1: Set Up the Workspace

To begin, open a web browser on your computer and navigate to the official tool: Teachable Machine . Click on the Get Started button, and then select Image Project from the options. Choose the Standard Image Model .

You will see a workspace with two default categories, which are called "classes" in machine learning. Have your child rename these classes. For this guide, let us use "Class 1: Mug" and "Class 2: Bottle". If they prefer, they can use "Thumbs Up" and "Thumbs Down".

Step 2: Gather the Training Data

Now, click the Webcam button under the first class. Have your child hold the first object (for example, a coffee mug) in front of the camera.

Instruct them to click and hold the Hold to Record button. While recording, they should slowly rotate the mug, move it closer to the camera, and move it further away. This ensures the computer captures the object from multiple angles. Aim to collect at least 150 to 200 image samples.

Repeat this exact process for the second class using the second object (for example, a water bottle). Make sure the background remains relatively consistent so the computer focuses on the object rather than the room.

Step 3: Train the Model

Once both classes have sufficient images, click the Train Model button in the middle column.

It is crucial to tell your child not to close or minimize the browser tab during this process, as the training happens directly inside the browser. The computer is now analyzing the pixels in the images, looking for patterns, colors, and shapes that distinguish the mug from the bottle. This step typically takes about ten to thirty seconds depending on your computer speed.

Step 4: Test and Break the Model

Once training is complete, the Preview panel on the right will activate, showing a live feed from the webcam. Have your child hold up the mug. The output bar at the bottom should show a high confidence level (near 100 percent) for "Mug". Now, have them hold up the bottle and observe the change.

To make this a true learning experience, encourage your child to try and "break" the model. What happens if they hold up a different mug that was not used in the training? What happens if they block half of the bottle with their hand? What happens if they hold up a completely different object, like a notebook? This demonstrates how the model generalizes based on the data it was given.

Parent & Educator Sidebar

Core AI Concept: Supervised Learning Supervised learning is the process of training an AI model using labeled data. In this project, the labeled data consisted of the images categorized under "Mug" and "Bottle". The computer does not actually know what a mug is; it simply recognizes mathematical patterns in the pixels associated with each label.

Conversation Starters:

• "Why do you think the computer got confused when we held up a different colored mug?" (Discusses data diversity and generalization).
• "How could we make this model smart enough to recognize any cup in our house?" (Discusses expanding the training dataset).
• "If we only trained the model with images of your hand, would it work for my hand?" (Introduces the concept of algorithmic bias).

Action Steps Summary:

• Open the Tool: Go to Teachable Machine and start a new Image Project.
• Create Classes: Name two distinct categories for the objects or gestures you want to classify.
• Record Samples: Capture at least 150 webcam images for each class from various angles and distances.
• Train and Test: Click train, then use the live preview to test the model's accuracy.
• Discuss the Concepts: Use the sidebar questions to help your child understand how data shapes AI behavior.

Looking for more family-friendly technology projects?