Overview
Watch a technical talk exploring the implementation of on-device learning capabilities for STM32 microcontrollers, presented by STMicroelectronics Research Scientist Beatrice Rossi and Politecnico di Milano PhD Student Michele Craighero. Learn about a new framework developed in C programming language that enables Convolutional Neural Networks (CNNs) training directly on microcontrollers, moving beyond the traditional approach of edge-only inference. Discover how this framework successfully personalizes a 1D-CNN for Human Activity Recognition, allowing models to adapt to specific users and changing conditions without external data sharing. Explore the accompanying software tool that helps estimate memory and computational requirements for model personalization, addressing key considerations in privacy preservation and real-time prediction capabilities for TinyML applications.
Syllabus
tinyML Talks: Enabling on-device learning on STM32 microcontrollers
Taught by
EDGE AI FOUNDATION