Teachable Machine: Teaching AI to School Students Without Coding

Artificial intelligence can feel abstract to children, and even to many teachers. The fastest way to make it real is to let students train a model themselves and watch it make predictions. Google’s free Teachable Machine does exactly that, with no coding at all. In this article we explain how it works and share practical Teachable Machine projects that bring AI for school students into an ordinary classroom with just a laptop and a webcam.

What Is Teachable Machine?

Teachable Machine is a web tool from Google that trains machine learning models directly in the browser. Students open the website, choose a project type, give examples, and click Train. Within a minute they have a working model they can test live.

It supports three kinds of projects:

  • Image project: recognises objects or gestures from webcam pictures or uploaded images.
  • Audio project: recognises sounds such as claps, whistles or spoken words from short microphone recordings.
  • Pose project: recognises body positions using keypoints on the body, such as arms raised or a squat.

Training happens in the browser, so schools do not need special servers. When a model is ready, students can export it. The tool offers TensorFlow.js export for use on websites, and image models can also be exported in formats for mobile and small devices. Older students can use the exported model in their own web pages or apps.

How Machine Learning for Kids Works in Teachable Machine

Every Teachable Machine project follows three steps, and these steps are the core idea of supervised machine learning:

  1. Gather: create classes (labels) and collect examples for each one. For example, “Apple” and “Banana”, with 50 to 100 pictures each.
  2. Train: the tool learns patterns that separate one class from another.
  3. Preview: show new examples and see the model’s confidence for each class as a percentage.

Because students collect the data themselves, they quickly see the most important lesson in AI: a model is only as good as the examples it learns from.

Five AI Activities for Kids in the Classroom

1. Fruit and Vegetable Sorter (Image)

Bring a few local items such as banana, tomato and onion. Students create one class per item and capture pictures from different angles. Then they test with items held in new positions. Discuss why the model sometimes gets confused when two items have similar colours. Suitable from Class 4 upwards.

2. Rock, Paper, Scissors (Image)

Students train three hand gesture classes plus a “Nothing” class for an empty frame. The extra class teaches a useful idea: models need a way to say “none of these”. Students can then play against a partner while the model acts as referee.

3. Clap or Whistle Detector (Audio)

Teachable Machine asks students to record background noise first. This is a good moment to explain that real systems must learn to ignore noise. Students then record claps, snaps and whistles. Test it in a quiet room and then in a noisy room, and compare the results.

4. Exercise and Yoga Pose Checker (Pose)

Students train poses like “arms up”, “standing” and “squat”, or simple yoga postures. This activity links AI with physical education and is very engaging for younger classes. It also shows how fitness apps and motion games track body positions.

5. Waste Segregation Assistant (Image)

Create classes for paper, plastic and food waste using clean sample items. Students discuss how such a model could help at a school recycling point. This connects AI to a real community problem and makes a strong science exhibition project.

Turning Activities into Real Learning

The click-and-train part is quick, so spend most of the period on discussion. These questions work well:

  • What happens if one class has 100 pictures and another has only 10?
  • Does the model still work with a different background or lighting?
  • Does it work equally well for every student’s hand or voice? If not, why?
  • Where do we see similar AI in daily life, such as phone cameras or voice assistants?

These questions introduce data quality, bias and testing in a way children understand. Teachers can also have students deliberately train a poor model with too few or one-sided examples, then fix it. Learning from a broken model is often more memorable than a perfect demo.

Practical Tips for Teachers

  • Use Chrome or another modern browser and allow camera and microphone access.
  • Work in groups of three or four, with one student handling the webcam.
  • Keep backgrounds plain while training, then test with busy backgrounds later.
  • Save projects to Google Drive or download the project file so work is not lost.
  • Avoid collecting images of students’ faces as a class dataset. Hands, objects and poses are enough to teach the ideas.

From No-Code AI to Coding AI

Teachable Machine is a starting point. After students understand classes, training and confidence, they are ready for the next steps: connecting models to apps, then moving to Python and computer vision in higher classes. This progression builds real understanding instead of memorised definitions.

Key takeaways

  • Teachable Machine trains image, sound and pose models in the browser with no coding.
  • Gather, train and preview mirror the real supervised learning process.
  • Simple activities teach big ideas like data quality, noise and bias.
  • Models can be exported to TensorFlow.js for use in web projects.
  • It works best as the first step towards Python and computer vision.

Bring Structured AI Learning to Your School

Teachable Machine is one of the tools in IRI’s Robotics & AI pathway for Class 1 to 12. Our grade-wise textbook series takes students from ScratchJr and Scratch through Arduino, MIT App Inventor and ESP32 to Python and computer vision, with weekly hands-on classes and real kits. The program is aligned with NEP 2020, and we support schools through the year. Book a free 1 to 2 hour demo class on your campus. Call 8606082634 or email inrobos@gmail.com.

Bring the IRI Robotics & AI pathway to your school

Weekly hands-on classes for Class 1–12 on IRI's own textbook series. Start with a free demo class on your campus.

See IRI’s robotics & AI classes for kids →

Call 8606082634 · inrobos@gmail.com

Contact IRI

Frequently asked questions

Is Teachable Machine free to use in schools?

Yes. Teachable Machine is a free web tool from Google. It runs in a browser, so a laptop with a webcam and microphone is enough.

From which class can students use Teachable Machine?

Students from around Class 4 can do guided image and pose activities. Older students can explore audio models and export their models for web projects.

Do students need to know coding for these AI activities?

No. Training and testing need no code. Coding becomes useful later, when students want to use an exported model in their own app or website.

Industrial Robotics Institute
Logo