Is it possible to deploy deep learning models on low-cost, low-power microcontrollers? While it may be surprising, the answer is a definite “yes”! In this talk, we’ll explain how the new TensorFlow Lite framework enables creating very lightweight DNN implementations suitable for execution on microcontrollers. We’ll illustrate how this works using an example of a 20 Kbyte DNN model that performs speech wake word detection, and will discuss how this generalizes to image-based use cases. We’ll introduce TensorFlow Lite, and will explore the key steps in implementing lightweight DNNs, including model design, data gathering, hardware platform choice, software implementation and optimization.