Course 2: Applications of TinyML
Chapter 3 · Course 2
Applications of TinyML
Real-world TinyML applications: keyword spotting, visual wake words, and anomaly detection. Covers TensorFlow Lite, quantization, data engineering, and responsible AI development.
Textbook companion: Vol I Ch 10–12 · Hardware Kits§3.1 Welcome to Applications of TinyML
| Topic | Type |
|---|---|
| Who's Who in TinyML2?! | Reading |
| Welcome to TinyML Applications | Slides |
| Building Blocks (from Course 1) | Slides |
| What You'll Learn in This Course | Slides |
| What Resources are Needed for this Course | Reading |
| Preview of TinyML Applications | Slides |
| The Role of Sensors in TinyML Applications | Reading |
| The Kit for Course 3 | Reading |
§3.2 AI Lifecycle and ML Workflow
| Topic | Type |
|---|---|
| ML Lifecycle Part 1 | Slides |
| ML Lifecycle Part 2 | Reading |
| ML Workflow Part 1 | Slides |
| ML Workflow Part 2 | Reading |
§3.3 ML on Mobile and Edge IoT Devices (Part 1)
| Topic | Type |
|---|---|
| TensorFlow: Where We Left Off | Reading |
| Introduction to TensorFlow Lite | Slides |
| Using the TFLite Converter in Colab | Colab |
| How to use TFLite Models | Reading |
| Running Models with TFLite in Colab | Colab |
| TFLite Optimizations and Quantization | Slides |
| TFLite Optimizations and Quantization in Colab | Colab |
| Quantization Aware Training | Slides |
| Quantization Aware Training Colab | Colab |
| Assignment: Quantization in TFLite | Colab |
| Assignment Solution | Reading |
§3.4 ML on Mobile and Edge IoT Devices (Part 2)
| Topic | Type |
|---|---|
| Why are 8-Bits Enough for ML? | Reading |
| Post Training Quantization (PTQ) | Slides |
| PTQ Weight Distribution Colab | Colab |
| Quantization Aware Training (QAT) | Slides |
| Inference Engine: TF vs. TFLite | Slides |
| Conversion and Deployment | Reading |
§3.5 Keyword Spotting
| Topic | Type |
|---|---|
| Introduction to Keyword Spotting (KWS) | Slides |
| Keyword Spotting Challenges/Constraints | Slides |
| Keyword Spotting Application Architecture Overview | Reading |
| Keyword Spotting Datasets | Slides |
| Keyword Spotting Dataset Creation | Reading |
| Keyword Spotting Data Collection / Pre-Processing | Slides |
| Spectrograms and MFCCs | Reading |
| Spectrograms and MFCCs in Colab | Colab |
| A Keyword Spotting Model | Slides |
| Keyword Spotting in Colab | Colab |
| Intro to Training in Colab | Slides |
| Training in Colab | Reading |
| Monitoring Training in Colab | Reading |
| Assignment: Training your own KWS Model | Colab |
| Assignment Solution | Reading |
| KWS Metrics | Slides |
| Streaming Audio | Slides |
| Cascade Architectures | Slides |
| Keyword Spotting in the Big Picture | Reading |
§3.6 Data Engineering for TinyML Applications
| Topic | Type |
|---|---|
| Introduction to Data Engineering | Reading |
| What's Data Engineering and Why It's Important | Slides |
| Dataset Standards: Speech Commands | Slides |
| Speech Commands Paper | Reading |
| Crowdsourcing Data for the Long Tail | Slides |
| Giving back to the Open Source Community | Reading |
| Reusing and Adapting Existing Datasets | Slides |
| Responsible Data Collection | Slides |
| Section Summary | Reading |
§3.7 Visual Wake Words
| Topic | Type |
|---|---|
| Introduction to Visual Wake Words (VWW) Application | Reading |
| What are Visual Wake Words (VWW)? | Slides |
| Visual Wake Words Challenges | Slides |
| Visual Wake Words Dataset | Slides |
| Data Privacy with Images | Reading |
| Neural Network Architectures for VWW | Slides |
| The Math Behind MobileNets Efficient Computation | Reading |
| Transfer Learning for VWW | Slides |
| Assignment: Transfer Learning in Colab | Colab |
| Assignment Solution | Reading |
| Common Myths and Pitfalls about Transfer Learning | Reading |
| Metrics for VWW | Slides |
| Section Summary | Reading |
§3.8 Anomaly Detection
| Topic | Type |
|---|---|
| Introduction to Anomaly Detection | Reading |
| What Is Anomaly Detection | Slides |
| Anomaly Detection in Industry | Slides |
| Industry 4.0 and TinyML | Reading |
| Anomaly Detection Datasets | Slides |
| MIMII Dataset | Reading |
| Real vs. Synthetic Data | Reading |
| Unsupervised Learning: K-Means in Colab | Colab |
| Unsupervised Learning with Autoencoders | Slides |
| Autoencoder Model Architecture | Reading |
| Training and Metrics for Autoencoders in Colab | Colab |
| Assignment: Training an Anomaly Detection Model | Colab |
| Assignment Solution | Reading |
| Section Summary | Reading |
§3.9 Responsible AI Development
| Topic | Type |
|---|---|
| Data Collection | Slides |
| The Many Faces of Bias in ML | Reading |
| Biased Datasets | Slides |
| Bias | Reading |
| Fairness | Slides |
| Google's What-If Tool in Colab | Colab |
| Fairness | 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.