Skip to main content
Planet Zephyr

Beefy ML: Ultra-Low Power Algorithms on Cattle – Jordan Yates, CSIRO

The Zephyr® Project strives to deliver the best-in-class RTOS for connected resource-constrained devices, built to be secure and safe. Click here to learn more: https://www.zephyrproject.org

Beefy ML: Ultra-Low Power Algorithms on Cattle – Jordan Yates, CSIRO

Machine learning may conjure images of GPU clusters in a datacenter, but robust algorithms can actually be run in microseconds on a basic Cortex M4F. Combined with thoughtful hardware design, this enables systems which can run ML algorithms indefinitely on the ear of an animal. In this session I will provide an overview of the journey to a real-world algorithm designed to monitor cattle feed intake from accelerometer data. The focus will be on the system design, algorithm implementation and validation, but data acquisition, annotation and training will also be touched on. The whole system is built upon multiple open-source components: * Zephyr as the underlying OS * Zephyr Power Management for low power operation * PyTorch for algorithm training * CMSIS DSP for embedded implementation

Benjamin Cabé