Tensorflow Lite for Microcontrollers Training Course

Introduction

• Microcontroller vs Microprocessor
• Microcontrollers designed for machine learning tasks

Overview of TensorFlow Lite Features

• On-device machine learning inference
• Solving network latency
• Solving power constraints
• Preserving privacy

Constraints of a Microcontroller

• Energy consumption and size
• Processing power, memory, and storage
• Limited operations

Getting Started

• Preparing the development environment
• Running a simple Hello World on the Microcontroller

Creating an Audio Detection System

• Obtaining a TensorFlow Model
• Converting the Model to a TensorFlow Lite FlatBuffer

Serializing the Code

• Converting the FlatBuffer to a C byte array

Working with Microcontroller'ss C++ Libraries

• Coding the microcontroller
• Collecting data
• Running inference on the controller

Verifying the Results

• Running a unit test to see the end-to-end workflow

Creating an Image Detection System

• Classifying physical objects from image data
• Creating TensorFlow model from scratch

Deploying an AI-enabled Device

• Running inference on a microcontroller in the field

Troubleshooting

Summary and Conclusion