Building upon our release last week of support for TensorFlow Lite for Microcontrollers, this week we’ve added a step-by-step tutorial showing how SensiML data collection, annotation, and feature preprocessing is combined with a neural network classifier using TFL Micro.

The tutorial is built around an example application and dataset for a smart boxing glove wearable integrating a Cortex-M class MCU with 6-axis accel+gyro IMU sensor to provide realtime punch recognition.

Fitting we believe, as the combination punch of SensiML and Google TensorFlow Lite for Microcontrollers should be a real knockout! <Ed note: Sorry I couldn’t resist.>

Original Post (from 9/16/20):

Today we pulled the covers off our latest update of which we are quite excited and suspect both SensiML and TFL users will be as well.

As of this week, SensiML Analytics Toolkit now provides pipeline support for one of the most popular open-source AI frameworks in existence: TensorFlow, specifically its TinyML variant TensorFlow Lite for Microcontrollers.

The combination of SensiML and TensorFlow Lite for Microcontrollers offers best-in-class AI code generation for TinyML applications from consumer wearables to multi-sensor industrial monitoring devices. SensiML Analytics Toolkit provides developers with a production-class tool for AI dataset management and AutoML preprocessing. Google’s TensorFlow Lite for Microcontrollers provides a popular well known subset of the TensorFlow framework for building and deploying neural network code on small microcontrollers.

To learn more about the integration and what it can do for your TinyML project, see the TensorFlow Lite Users page.

* All product names, logos, and brands are property of their respective owners. 


“The integration with SensiML Analytics Studio and Google’s TensorFlow Lite for Microcontrollers provides a nice end-to-end workflow for creating IoT edge models on embedded devices”

Ian Nappier, TensorFlow Lite for Microcontrollers Product Manager at Google