Teachable Machine
3 Minutes · Module 3 - Tools for ML
Teachable Machine tutorial for students explains how Google's web tool helps train image, audio, and pose models. A Tamil technology lesson.
- Three steps: uploading data, training the model, and testing it
- Standard versus embedded image models and their uses
- Advanced training settings: epochs, batch size, and learning rate
What this lesson teachesRead the lesson as text
Teachable Machine is a web-based platform developed by Google that allows students and educators to build machine learning models quickly without coding. It runs directly inside web browsers such as Chrome or Safari and provides a simple way to train models using images, audio, or poses.
The three core steps
Creating a working machine learning model on the platform involves three simple steps. First, users upload data samples into separate categories called classes, such as Class 1 and Class 2. Next, clicking the train model button allows the system to analyze and learn from the uploaded data. Finally, users test the trained model in the preview section by giving it new inputs to check its results.
Standard versus embedded models
When starting an image project, the platform offers standard and embedded image model options. Standard image models are mainly used for digital software like websites, mobile apps, and smartphone face unlock features. Embedded image models are designed for hardware automations, where machine learning models are sent to microcontrollers to operate physical actuators like motors and lights.
Advanced training settings
- Epochs define the number of times the system analyzes the uploaded samples, and setting a higher number increases accuracy.
- Batch size organizes the uploaded data into specific groups, such as sets of sixteen or thirty-two images.
- Learning rate controls how the machine learns from its outputs, where a lower rate learns slowly without errors, and a higher rate learns faster with a risk of mistakes.
Supported project types and uses
The tool supports three main types of projects, allowing users to train image models, audio models, and pose models. After checking the model in the preview panel, the resulting system can be put to work in different environments. These include digital platforms like websites and apps, as well as electronic hardware powered by microcontrollers.
Part of the course
A.I. with Teachable Machine