Open-Source AutoML Enhances Edge AI Development with Kenning Framework Integration

Analog Devices and Antmicro have released AutoML for Embedded, an open-source tool integrated into the Kenning framework, designed to streamline the development and deployment of machine learning models on edge devices like microcontrollers. This tool addresses the challenges developers face in managing data preprocessing, model selection, hyperparameter tuning, and hardware-specific optimizations for resource-constrained platforms.

AutoML for Embedded automates the machine learning pipeline, enabling developers to create lightweight, efficient models tailored for edge devices. It is available as a Visual Studio Code plugin and supports ADI’s MAX78002 AI Accelerator MCUs and MAX32690, as well as Renode-based simulation and Zephyr RTOS for prototyping and testing. The tool also supports general-purpose, open-source tools to avoid platform lock-in and includes tutorials, reproducible pipelines, and example datasets for accessibility.

The tool leverages SMAC (Sequential Model-based Algorithm Configuration) for exploring model architectures and training parameters, and Hyperband with Successive Halving to prioritize promising models. It verifies model size against device RAM to ensure deployment feasibility. Models can be optimized, evaluated, and benchmarked using Kenning’s flows, with reports on size, speed, and accuracy.

A demonstration showed...