CEC – Machine Learning Application Design using STM32 MCUs

Machine learning is quickly finding applications in the microcontroller space. Machine learning is a completely different paradigm in application development. In this course, we will explore how to get starting with machine learning on a microcontroller and how to meld it with the traditional application design methodologies.

Registration and Playback located here (May require login to access)

April 26 – Day 1 – Introduction to Machine Learning on MCU’s Machine learning (ML) has often been considered a technology that operates on high-end servers and doesn’t have a place in traditional embedded systems. That perception is quickly changing. In this session, attendees will get a brief introduction to machine learning and how it is being leveraged on microcontroller-based devices.

April 27 – Day 2 – Capturing, Cleaning and Labeling Data

Training a ML model requires that a developer capture, clean and label data. This requires a developer to not only carefully select their dataset, but also figure out how it will be processed on the target. In this session, we will explore how to identify, capture, clean and perform digital signal processing on the data prior to building an ML model.

April 28 – Day 3 – Training A Neural Network Part 1

Once developers have the data they want to use and have captured and labeled an adequate dataset, its time to train a model. In this session, we will start to investigate the tools developers can use to build a machine learning model. We will explore tools such as TensorFlow Lite and STM32CubeMx.

April 29 – Day 4 – Training a Neural Network Part 2

Training a network isn’t a trivial endeavor. It often requires as much art as it does science in order to be successful. In this session, we will continue to explore how to train a model. We will discuss how to test the model and determine how accurate it is. Attendees will walk away with an understanding of how to use the STM32 toolchain to simplify training and deploying their model.

April 29 – Day 5 –Running an Inference on Target

Creating a model and testing it is great, but it’s not complete until the inference is deployed on a target and ran in the real world. In this session, we will examine how to deploy a model onto an STM32 development board and integrate it with existing application code. Attendees will learn how to have their inference co-exist with traditional embedded software.

Course Resources

Jacob’s General Embedded System Resources:

  • Sign-Up for the Embedded Bytes Newsletter here
  • Developing Reusable Firmware – A Practical Guide to API’s, HAL’s and Drivers here
  • MicroPython Projects Book here
  • Jacob’s YouTube Channel – here

Course Resources:

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.