Teachable Machine

Train custom image classification models by uploading or capturing images from a webcam, organizing them into labeled classes, and training a model in the browser without writing code.

Train custom audio classification models by recording or uploading sound samples, labeling them by category, and generating a model that recognizes audio patterns without coding.

Upload pre-existing image or audio files from a device's file system to use as training data, supporting batch imports of multiple files organized by class.

Organize training samples into labeled classes or categories, with visual feedback showing sample counts and distribution across classes to ensure balanced training data.

Train and export models without requiring account creation or login, enabling immediate access to the platform for casual users and classroom demonstrations.

Display prediction results with confidence scores showing how certain the model is about each classification, providing transparency into model decision-making.

Train custom pose recognition models by capturing body positions via webcam, labeling different poses or movements, and creating a model that identifies human body positions and gestures.

Test trained models in real-time within the browser interface using live webcam feed or uploaded samples, with immediate visual feedback on predictions and confidence scores.

Export trained models in TensorFlow.js format for deployment in web applications, enabling models to run directly in browsers without server-side infrastructure.

Export trained models in TensorFlow SavedModel format for deployment on servers, mobile devices, or edge devices using TensorFlow Lite or other production frameworks.

Provide a drag-and-drop, visual interface for training machine learning models without requiring any programming knowledge, code writing, or technical ML expertise.

Train models that can classify inputs into multiple distinct categories or classes, allowing users to create complex classification systems by organizing training data into labeled groups.

Execute all model training computations locally within the web browser using TensorFlow.js, eliminating the need for cloud uploads, server infrastructure, or external processing.

Capture training data directly from a device's webcam in real-time, allowing users to quickly gather image or pose samples without needing external files or equipment.