Course 3: Deploying TinyML
Chapter 4 · Course 3
Deploying TinyML
End-to-end deployment with TensorFlow Lite for Microcontrollers. Covers embedded hardware/software, TFLite Micro internals, and hands-on labs deploying keyword spotting, visual wake words, and gesture recognition on Arduino.
Textbook companion: Vol I Ch 11, 13 · Arduino & Seeed Kits§4.1 Welcome to Deploying TinyML
| Topic | Type |
|---|---|
| Welcome to TinyML3 | Reading |
| Welcome Message from VJ | Slides |
| Course 1 and 2 Recap | Reading |
| TinyML Application Deployment Preview | Slides |
| The TinyML Kit | Reading |
| TinyML Course Kit Overview | Slides |
| How the Course is Structured | Slides |
§4.2 Getting Started
| Topic | Type |
|---|---|
| C++ for Python Users | Reading |
| Setting up your Hardware | Reading |
| Setting up your Software | Reading |
| The Arduino Blink Example | Reading |
| Testing the TensorFlow Install | Reading |
| Testing the Sensors | Reading |
§4.3 Embedded Hardware and Software
| Topic | Type |
|---|---|
| Embedded System | Slides |
| Diversity of Embedded Systems | Reading |
| Embedded Computing Hardware | Slides |
| Diversity of Embedded Microcontrollers | Reading |
| Embedded I/O | Slides |
| Transducer Modules and Wireless Communication | Reading |
| Embedded System Software | Slides |
| Arduino cores, frameworks, mbedOS, and 'bare metal' | Reading |
| Embedded ML Software | Slides |
§4.4 TensorFlow Lite Micro
| Topic | Type |
|---|---|
| What is TensorFlow Lite for Microcontrollers? | Reading |
| TFMicro: The Big Picture | Slides |
| TFLite Micro: Interpreter | Slides |
| MCU Memory Hierarchy | Reading |
| TFLite Micro: Model Format / FlatBuffer | Slides |
| TensorFlow Lite Flatbuffer Manipulation Colab | Colab |
| TFLite Micro: Memory Allocation (Tensor Arena) | Slides |
| TFLite Micro: NN Operator Support (OpsResolver) | Slides |
| TFLite Micro Developer Design Principles | Reading |
§4.5 Keyword Spotting
| Topic | Type |
|---|---|
| TinyML "Keyword Spotting" Workflow | Reading |
| KWS Application Architecture | Slides |
| KWS Initialization | Slides |
| KWS Pre-processing | Slides |
| KWS Inference | Slides |
| KWS Post-processing | Slides |
| KWS Summary | Slides |
§4.6 Custom Dataset Engineering for Keyword Spotting
| Topic | Type |
|---|---|
| Recap Dataset Engineering | Reading |
| Introducing Custom Dataset for KWS | Slides |
| Things to Consider for Your Data Collection Plan | Reading |
| Building a Custom Dataset | Reading |
| Train and Deploy Your Custom Dataset KWS Model | Reading |
§4.7 Visual Wake Words
| Topic | Type |
|---|---|
| Recap: What are Visual Wake Words? | Reading |
| Person Detection Application Architecture | Slides |
| Person Detection with KWS: MultiModal | Slides |
| Person Detection with KWS: MultiTenancy | Slides |
| MultiTenancy in TensorFlow Lite Micro | Slides |
| Deploying the Pretrained Person Detection Model | Reading |
| Deploying a Multi-Tenant Application | Reading |
§4.8 Gesture Recognition — Magic Wand
| Topic | Type |
|---|---|
| Recap: Time Series for Anomaly Detection | Reading |
| TinyML Sensor Ecosystem | Slides |
| Anatomy of an IMU | Reading |
| Magic Wand Application | Slides |
| Magic Wand Application Architecture | Slides |
| Understanding the Magic Wand Application | Reading |
| Deploying the Magic Wand | Reading |
| Collecting Data for Your Custom Magic Wand Project | Reading |
| Training and Deploying Your Custom Magic Wand Project | Reading |
§4.9 Responsible AI Deployment
| Topic | Type |
|---|---|
| Privacy | Slides |
| Privacy | Reading |
| Security | Slides |
| Attacking a KWS Model in Colab | Colab |
| Why do ML Models Fail after Deployment? | Reading |
| Monitoring after Deployment | Slides |
§4.10 Summary
| Topic | Type |
|---|---|
| Congratulations! You Made it to the Finish Line! | Slides |
| What Comes Next: Advanced Topics in TinyML | Reading |
| What Do I Do Now? | Slides |
| TinyMLx Project Extension (Optional) | Reading |
Note
These materials were originally developed for the HarvardX Professional Certificate in Tiny Machine Learning on edX. See the original curriculum for the full item-by-item breakdown including forum prompts and quizzes not listed above.