edge-ml

edge-ml is an open-source, browser-based toolchain designed to bring machine learning directly onto microcontrollers. Developed at TECO (KIT), it simplifies the entire lifecycle of embedded machine learning from data collection to deployment by providing intuitive libraries, cloud integration, and model optimization for edge devices.

The main goal of edge-ml is to lower the barrier for embedded machine learning. It aims to enable rapid prototyping, efficient deployment, and widespread adoption of intelligent applications on microcontrollers by simplifying the complete development workflow.

• Streamlined Workflow: Allows users to record data, label samples, train models, validate results, and deploy embedded ML applications with minimal effort via a web-based tool.
• Automated Optimization (edge-ml AUTO): An automated neural architecture search feature that delivers highly optimized models for embedded hardware. It achieved 95.9% accuracy with just 7008 B peak memory on the UCI-HAR dataset, significantly outperforming TensorFlow Lite.
• Broad Platform Support: Supports a wide range of frameworks (Arduino, Node.js, Python) and edge devices (Arduino Nicla Sense ME, Nano 33 BLE, ESP32).

By making embedded machine learning more accessible, edge-ml enables rapid innovation in edge AI. Its combination of an easy-to-use toolchain, automated optimization, and strong community support (Discord and GitHub) makes it a powerful framework for developers building intelligent applications on resource-constrained devices.