When: Wed 26.02.2025 12:00 - 17:00
Where: KYPO - S108, Fakulta informatiky, Botanická 68a
Registration: https://forms.office.com/e/gJRsMFBgk1
open for everybody
Questions: mailto:jan.kral@fi.muni.cz

Duration: approx. 45 min

• STM32 MCUs Portfolio Update & ST MEMS Portfolio Update

Join us for a hands-on training session focused on rapidly developing and integrating machine learning (ML) models within embedded applications using resource-constrained microcontrollers (MCUs), such as the Cortex-M0+ with 32 Kbytes RAM and 256 Kbytes flash. This training is designed to accelerate model development through automated tools, minimizing the need for extensive data science expertise. Join us to experience the practicality and potential of ML in embedded systems!

Key components of the class: Duration: approx. 3 hours

• Anomaly detection use case: Explore predictive maintenance for mechanical parts, specifically linear motors. Due to practical constraints, attendees will emulate motor movements by hand.

• Relevant components: Utilize MEMS sensors like accelerometers and gyroscopes for movement trajectory prediction.

• Comprehensive development pipeline:
  
  • Data collection: Gather input datasets.
    
  • ML model selection and benchmarking: Choose and evaluate the best models.
    
  • Model emulation: Simulate ML model behavior using real-time data streams.
    
  • Model library generation: Create and integrate ML model libraries within embedded applications.

Expected outcomes

Develop ML model libraries with a binary size of a few Kbytes of RAM and flash, capable of learning on the target MCU during runtime. Our goal is to ignite a passion for efficient algorithm coding and demonstrate that ML is feasible on limited-resource MCUs, including runtime learning on the target MCU.

General discussion about ST MCUs and Sensors.

Presentation and templates:
https://owncloud.cesnet.cz/index.php/s/8FEN8GD6CwN8337

Teams Link:
https://teams.microsoft.com/meet/332156999227?p=aoxBJGb7a5QaF5H2tl