Recognizing words on a microcontroller using TinyML

AI on IoT is moving from the cloud to the edge, running models closer to the data. Traditionally the hardware to run these models on the edge has been powerful, with GPUs or compute sticks. But what if you could run a model in only a few kilobytes of memory on a tiny micro-controller drawing less than a milliwatt of power? In this video we look at doing just that, training a wake word model in the cloud using Azure ML Studio, then compressing it to 18KB and running it on an Adafruit EdgeBadge, a small, low-powered micro-controller based device.

Jump To: 

• [02:00] – The history of IoT
• [04:25] – Intro to TinyML
• [05:12] – Intro to the EdgeBadge
• [06:04] – Training the model
• [08:03] – Demoing the notebook
• [10:12] – Using VS Code to program the EdgeBadge
• [11:41] – Demo the wake word

Learn More: 

• Adafruit Edge Badge
• All the Conent you need to Create a sample

The AI Show's Favorite links:

• Don't miss new episodes, subscribe to the AI Show  
• Create a Free account (Azure) 
• Deep Learning vs. Machine Learning   
• Get Started with Machine Learning