Amber - A Complete, ML-Based, Anomaly Detection Pipeline for Microcontrollers

Explore a complete, unsupervised machine learning-based anomaly detection pipeline deployable on low-power microcontrollers in this tinyML Talks webcast. Discover how the Amber algorithm seamlessly tunes hyperparameters, trains ML models, and transitions to anomaly detection mode using live sensor values in real-time. Learn about its ability to generate thousands of inferences per second with high accuracy on ARM Cortex M7 processors. Gain insights into sensor anomaly detection pipelines, including data collection, off-line training, and model-building processes. Compare traditional static models with Amber's dynamic approach, which can adapt to various sensors and assets. Delve into topics such as analytics, synthetic data, epilepsy applications, benchmarks, and the Boon Nano technology. Hear from speakers Brian Turnquist and Rodney Dockter of Boon Logic as they present this innovative solution for microcontroller-based anomaly detection.