Build a Teachable Machine Image Classifier With Your Kid

Introduce your child to the fundamentals of machine learning through a hands-on, interactive project using a free web tool and a standard webcam.

Most children interact with artificial intelligence daily, whether through video game matchmaking, video recommendation algorithms, or voice assistants. However, they rarely get to see behind the curtain to understand how these systems actually learn. By building a custom image classifier, you can shift your child from a passive consumer of technology to an active creator. Using a free, web-based tool developed by Google, you and your child can train a computer to recognize hand gestures, household objects, or facial expressions in less than twenty minutes. This hands-on experience demystifies the core concepts of machine learning, providing a solid foundation for their digital future.

The Project Setup

To begin this project, you do not need any paid subscriptions, specialized software, or coding experience. All that is required is a computer or tablet equipped with a working webcam and an internet connection. This accessibility makes it an ideal weekend activity for children ages 8 to 16, allowing them to engage directly with cutting-edge technology.

Start by opening a web browser and navigating directly to the Google Teachable Machine website at https://teachablemachine.withgoogle.com/ to access the platform. Once the page loads, click on the Get Started button. You will be presented with three project options: Image Project, Audio Project, and Pose Project. Select the Image Project option, and then click on Standard Image Model. This will open the main workspace where you and your child will build, train, and test your custom artificial intelligence model.

Gathering and Labeling Training Data

The first step in training any machine learning model is gathering data. In this project, you and your child will teach the computer to recognize three distinct hand gestures: a Thumbs Up, a Thumbs Down, and a Neutral state where no gesture is shown. This simple setup provides a clear, visual way to understand how computers categorize information.

In the workspace, you will see two default boxes labeled Class 1 and Class 2. Click the pencil icon next to Class 1 and rename it to Thumbs Up. Click the webcam button within that box to activate your camera. Instruct your child to hold their hand in front of the camera, showing a clear thumbs-up gesture. Click and hold the Record button to capture approximately one hundred image samples. Encourage your child to slowly tilt their hand, move it closer to the lens, and shift it slightly while recording. This variation helps the model learn what a thumbs-up looks like from different angles.

Next, rename Class 2 to Thumbs Down. Repeat the recording process, having your child hold a clear thumbs-down gesture in front of the camera while capturing another one hundred samples. Finally, click the Add a Class button at the bottom of the list to create a third category. Rename this class to Neutral. For this class, record your child sitting quietly in front of the camera without making any hand gestures. This neutral class is crucial because it teaches the computer what the background looks like when no active gesture is being performed.

Training the Model

With your training data gathered and labeled, you are ready to train your model. In the middle column of the workspace, you will see a box labeled Training with a blue button that says Train Model. Click this button to begin the process. Behind the scenes, the browser-based algorithm is analyzing the hundreds of images you just captured. It is looking for patterns, such as the direction of the thumb, the shape of the hand, and the contrast against the background, to establish mathematical rules for each class.

This training process takes between ten and thirty seconds, depending on the speed of your computer. It is important to keep the browser tab open and active during this time. Instruct your child to watch the progress bar. This short wait is an excellent opportunity to explain that the computer is actively learning from the examples they provided, much like a student studying flashcards before an exam.

Testing and Iterating

Once the training is complete, the right-hand column will activate, showing a live preview from your webcam. Below the preview, you will see three progress bars corresponding to your classes: Thumbs Up, Thumbs Down, and Neutral. As your child makes different gestures in front of the camera, the progress bars will shift in real time, showing the model's confidence level for each gesture.

This is where the real learning happens. Encourage your child to test the limits of the model. What happens if they use their left hand instead of their right hand? What happens if they make the gesture very far away from the camera? Often, the model will make mistakes, such as misclassifying a left-handed thumbs-up as neutral. Explain to your child that this is not a failure of the computer, but rather a limitation of the data they provided. To fix these errors, simply go back to the corresponding class, record additional samples using the left hand, and click Train Model again. This iterative process teaches children that artificial intelligence is only as unbiased and accurate as the data used to train it.

Parent and Educator Sidebar

Core AI Concept: Supervised Learning Supervised learning is a type of machine learning where a model is trained on labeled data. In this project, the labels were Thumbs Up, Thumbs Down, and Neutral, and the data was the collection of webcam images. The computer learns by finding common patterns among the images in each labeled group, allowing it to make predictions when shown new, unlabeled images.

Conversation Starters:

• How do you think a self-driving car uses this type of technology to recognize the difference between a stop sign, a pedestrian, and a green light?
• Why did the computer get confused when we changed hands or moved far away, and what does that tell us about how careful we need to be when collecting data?
• Can you think of another project we could build using this tool, such as training the computer to recognize different family pets or sorting recyclable items from trash?

Action Steps Summary

• Open a web browser and go to the Google Teachable Machine website at https://teachablemachine.withgoogle.com/ to start.
• Select Image Project and choose Standard Image Model to open the workspace.
• Create and label three classes: Thumbs Up, Thumbs Down, and Neutral.
• Use the webcam to record one hundred to one hundred and fifty diverse image samples for each class.
• Click Train Model and keep the browser tab active while the algorithm processes the data.
• Test the model using the live preview and iterate by adding more diverse images to correct any errors.

Ready to inspire the next generation of creators?