Build a Teachable Machine Image Classifier With Your Kid

Published January 17, 2024

Children today are surrounded by artificial intelligence, from the recommendation algorithms on YouTube to the voice assistants in their living rooms. However, most kids experience AI as passive consumers, viewing it as a form of digital magic rather than a tool built on logic and data. For busy executives, finding high-quality, educational activities to share with your children can be challenging, especially when trying to balance screen time with meaningful learning.

The best way to demystify AI is to have your child build a working model themselves. You do not need a computer science degree, expensive hardware, or a paid software subscription to make this happen. Using a free, web-based tool developed by Google, you and your child can train a custom image classifier in less than twenty minutes. This hands-on project transforms abstract concepts into concrete understanding, giving your child a foundational grasp of how modern machine learning works.

Demystifying Machine Learning

To get started, you only need a computer with a functional webcam and an internet connection. There are no accounts to create, no software to install, and no credit cards required. This accessibility makes it an ideal weekend activity for children aged 8 to 16, allowing them to focus entirely on the creative and logical aspects of the technology.

The platform we will use is Google's Teachable Machine, a web-based tool that allows users to train machine learning models quickly and intuitively. You can access the tool directly at https://teachablemachine.withgoogle.com/ . This tool provides a visual interface where children can see the direct relationship between the data they input and the decisions the computer makes.

For this project, we will build a classic Rock, Paper, Scissors game classifier. The goal is to train the computer to recognize which of the three hand gestures your child is showing to the webcam. This project is highly interactive and provides immediate visual feedback, making it perfect for keeping younger minds engaged.

Step-by-Step Guide to Building Your Classifier

Step 1: Set Up Your Classes. When you open Teachable Machine, select Image Project and then choose Standard Image Model. You will see a workspace with two default classes. Think of classes as categories or folders that the computer uses to group information. Rename Class 1 to Rock and Class 2 to Paper. Click the Add a class button to create a third category and name it Scissors.

Step 2: Gather Your Training Data. This is where the hands-on fun begins. Click the Webcam button under the Rock class. Have your child hold their hand in a fist (the Rock gesture) in front of the webcam. Press and hold the Hold to Record button. Encourage your child to move their hand slightly, tilt it, bring it closer to the camera, and move it further away while recording. This variety helps the computer learn what a Rock looks like from different angles. Collect about 100 to 150 image samples. Repeat this exact process for the Paper class (flat hand) and the Scissors class (two fingers extended).

Step 3: Train Your Model. Once you have gathered data for all three classes, click the Train Model button in the middle column. This process takes about ten to thirty seconds. Explain to your child that the computer is currently looking for patterns in the images, such as the shape of the hand, the position of the fingers, and the contrast against the background. Warn them not to close the browser tab while the training is in progress.

Step 4: Test and Play. Once training is complete, the Preview column on the right will activate, showing a live feed from your webcam. Have your child make one of the three gestures in front of the camera. Below the video feed, they will see three progress bars representing the classes: Rock, Paper, and Scissors. The bars will dynamically shift to show the percentage of confidence the computer has in its prediction. If your child shows a fist, the Rock bar should shoot up to 100 percent.

Exploring Model Limitations and Bias

Once the basic model is working, it is time to experiment and test its limits. This is where the deepest learning occurs. Ask your child to step out of the frame and have you, the parent, show a Rock gesture. Does the computer still recognize it? What happens if you use your left hand instead of your right hand? What happens if you change the lighting in the room or stand against a different background?

If the computer struggles to recognize your hand, explain that this is due to a lack of diverse training data. If your child only trained the model using their own right hand in bright lighting, the computer only knows how to recognize that specific scenario. To fix this, you can record additional images of your hand, or record in different lighting, and retrain the model. This simple exercise teaches children a profound lesson about algorithmic bias and the importance of diverse data sets in real-world AI applications.

Parent & Educator Sidebar

Core AI Concept: Supervised Learning

Supervised learning is a type of machine learning where the computer learns from labeled training data. In this project, the labeled data consists of the images categorized under Rock, Paper, and Scissors. The computer does not inherently know what a hand is, it simply identifies mathematical patterns in the pixels of the images associated with each label.

Conversation Starters:

• How do self-driving cars use this? How does a self-driving car tell the difference between a stop sign, a pedestrian, and a plastic bag blowing in the wind? How much training data do you think they need?
• What is algorithmic bias? If we only trained our model with small hands, why did it struggle to recognize a parent's larger hand? How could this cause problems in real-world systems like facial recognition?
• How can we trick the machine? Can you make a gesture that looks halfway between Rock and Paper? How does the computer react when it is confused?

Action Steps Summary

• Visit Teachable Machine: Go to https://teachablemachine.withgoogle.com/ and start a standard image project.
• Create Three Classes: Label them Rock, Paper, and Scissors.
• Record Training Data: Use your webcam to capture 100 to 150 images of each hand gesture from various angles.
• Train the Model: Click the train button and wait for the browser to process the images.
• Test and Refine: Use the live preview to test the model, identify where it fails, and add more diverse images to improve its accuracy.

Want more practical AI workflows for your daily routine?