Machine Learning on Microcontrollers - Tiny, Useful and Fun

Explore the world of TinyML through this 25-minute conference talk that introduces machine learning applications on microcontroller-grade systems. Learn about the practical constraints and opportunities of implementing ML on devices with under 1 watt power consumption, under 1 MB of RAM and FLASH memory, and under $10 bill-of-materials cost. Discover how TinyML differs from resource-intensive generative AI models by focusing on sensor data analysis in embedded systems. Gain insights into recent developments in the field and understand various application areas where TinyML provides valuable solutions. Get practical guidance on starting your own TinyML projects using the emlearn open-source project, which offers machine learning inference implementations written in portable C99 code. Understand how emlearn supports models trained with popular frameworks like scikit-learn and TensorFlow/Keras, and learn about its MicroPython support introduced in 2023 for Python-based microcontroller development.