Embedded Machine Learning Courseware

Welcome to the Edge Impulse open courseware for embedded machine learning! This repository houses a collection of slides, reading material, project prompts, and sample questions to get you started creating your own embedded machine learning course. You will also have access to videos that cover much of the material. You are welcome to share these videos with your class either in the classroom or let students watch them on their own time.

This repository is part of the Edge Impulse University Program. Please see this page for more information on how to join: edgeimpulse.com/university.

How to Use This Repository

Please note that the content in this repository is not intended to be a full semester-long course. Rather, you are encouraged to pull from the modules, rearrange the ordering, make modifications, and use as you see fit to integrate the content into your own curriculum.

For example, many of the lectures and examples from the TinyML Courseware (given by [3]) go into detail about how TensorFlow Lite works along with advanced topics like quantization. Feel free to skip those sections if you would just like an overview of embedded machine learning and how to use it with Edge Impulse.

In general, content from [3] cover theory and hands-on Python coding with Jupyter Notebooks to demonstrate these concepts. Content from [1] and [2] cover hands-on demonstrations and projects using Edge Impulse to deploy machine learning models to embedded systems.

Content is divided into separate modules. Each module is assumed to be about a week's worth of material, and each section within a module contains about 60 minutes of presentation material. Modules also contain example quiz/test questions, practice problems, and hands-on assignments.

If you would like to see more content than what is available in this repository, please refer to the Harvard TinyMLedu site for additional course material.

License

Unless otherwise noted, slides, sample questions, and project prompts are released under the Creative Commons Attribution NonCommercial ShareAlike 4.0 International (CC BY-NC-SA 4.0) license. You are welcome to use and modify them for educational purposes.

The YouTube videos in this repository are shared via the standard YouTube license. You are allowed to show them to your class or provide links for students (and others) to watch.

Professional Development

Much of the material found in this repository is curated from a collection of online courses with permission from the original creators. You are welcome to take the courses (as professional development) to learn the material in a guided fashion or refer students to the courses for additional learning opportunities.

• Introduction to Embedded Machine Learning - Coursera course by Edge Impulse that introduces neural networks and deep learning concepts and applies them to embedded systems. Hands-on projects rely on training and deploying machine learning models with Edge Impulse. Free with optional paid certificate.
• Computer Vision with Embedded Machine Learning - Follow-on Coursera course that covers image classification and object detection using convolutional neural networks. Hands-on projects rely on training and deploying models with Edge Impulse. Free with optional paid certificate.
• Tiny Machine Learing (TinyML) - EdX course by Vijay Janapa Reddi, Laurence Moroney, Pete Warden, and Lara Suzuki. Hands-on projects rely on Python code in Google Colab to train and deploy models to embedded systems with TensorFlow Lite for Microcontrollers. Paid course.

Prerequisites

Students should be familiar with the following topics to complete the example questions and hands-on assignments:

• Algebra
  • Solving linear equations
• Probability and Statistics
  • Expressing probabilities of independent events
  • Normal distributions
  • Mean and median
• Programming
  • Arduino/C++ programming (conditionals, loops, arrays/buffers, pointers, functions)
  • Python programming (conditionals, loops, arrays, functions, NumPy)

Optional prerequisites: many machine learning concepts can be quite advanced. While these advanced topics are briefly discussed in the slides and videos, they are not required for quiz questions and hands-on projects. If you would like to dig deeper into such concepts in your course, students may need to be familiar with the following:

• Linear algebra
  • Matrix addition, subtraction, and multiplication
  • Dot product
  • Matrix transposition and inversion
• Calculus
  • The derivative and chain rule are important for backpropagation (a part of model training)
  • Integrals and summation are used to find the area under a curve (AUC) for some model evaluations
• Digital signal processing (DSP)
  • Sampling rate
  • Nyquist–Shannon sampling theorem
  • Fourier transform and fast Fourier transform (FFT)
  • Spectrogram
• Machine learning
  • Logistic regression
  • Neural networks
  • Backpropagation
  • Gradient descent
  • Softmax function
  • K-means clustering
• Programming
  • C++ programming (objects, callback functions)
  • Microcontrollers (hardware interrupts, direct memory access, double buffering, real-time operating systems)

Feedback and Contributing

If you find errors or have suggestions about how to make this material better, please let us know! You may create an issue describing your feedback or create a pull request if you are familiar with Git.

This repo uses automatic link checking and spell checking. If continuous integration (CI) fails after a push, you may find the dead links or misspelled words, fix them, and push again to re-trigger CI. If dead links or misspelled words are false positives (i.e. purposely malformed link or proper noun), you may update .mlc_config.json for links to ignore or .wordlist.txt for words to ignore.

Required Hardware and Software

Students will need a computer and Internet access to perform machine learning model training and hands-on exercises with the Edge Impulse Studio and Google Colab. Students are encouraged to use the Arduino Tiny Machine Learning kit to practice performing inference on an embedded device.

A Google account is required for Google Colab.

An Edge Impulse account is required for the Edge Impulse Studio.

Students will need to install the latest Arduino IDE.

Preexisting Datasets and Projects

This is a collection of preexisting datasets, Edge Impulse projects, or curation tools to help you get started with your own edge machine learning projects. With a public Edge Impulse project, note that you can clone the project to your account and/or download the dataset from the Dashboard.

Motion

• Edge Impulse continuous gesture (idle, snake, up-down, wave) dataset
• Alternate motion (idle, up-down, left-right, circle) project

Sound

• Running faucet dataset
• Google speech commands dataset
• Keyword spotting dataset curation and augmentation script
• Multilingual spoken words corpus

Image Classification

• Electronic components dataset
• Image data augmentation script

Object Detection

• Face detection project

Syllabus

• Module 1: Machine Learning on the Edge
  • Learning Objectives
  • Section 1: Machine Learning on the Edge
  • Section 2: Limitations and Ethics
  • Section 3: Getting Started with Colab
• Module 2: Getting Started with Deep Learning
  • Learning Objectives
  • Section 1: Machine Learning Paradigm
  • Section 2: Building Blocks of Deep Learning
  • Section 3: Embedded Machine Learning Challenges
• Module 3: Machine Learning Workflow
  • Learning Objectives
  • Section 1: Machine Learning Workflow
  • Section 2: Data Collection
  • Section 3: Model Training and Evaluation
• Module 4: Model Deployment
  • Learning Objectives
  • Section 1: Quantization
  • Section 2: Embedded Microcontrollers
  • Section 3: Deploying a Model to an Arduino Board
• Module 5: Anomaly Detection
  • Learning Objectives
  • Section 1: Introduction to Anomaly Detection
  • Section 2: K-means Clustering and Autoencoders
  • Section 3: Anomaly Detection in Edge Impulse
• Module 6: Image Classification with Deep Learning
  • Learning Objectives
  • Section 1: Image Classification
  • Section 2: Convolutional Neural Network (CNN)
  • Section 3: Analyzing CNNs, Data Augmentation, and Transfer Learning
• Module 7: Object Detection and Image Segmentation
  • Learning Objectives
  • Section 1: Introduction to Object Detection
  • Section 2: Image Segmentation and Constrained Object Detection
  • Section 3: Responsible AI
• Module 8: Keyword Spotting
  • Learning Objectives
  • Section 1: Audio Classification
  • Section 2: Spectrograms and MFCCs
  • Section 3: Deploying a Keyword Spotting System

Course Material

Module 1: Machine Learning on the Edge

This module provides an overview of machine learning and how it can be used to solve problems. It also introduces the idea of running machine learning algorithms on resource-constrained devices, such as microcontrollers. It covers some of the limitations and ethical concerns of machine learning. Finally, it demonstrates a few examples of Python in Google Colab, which will be used in early modules to showcase how programming is often performed for machine learning with TensorFlow and Keras.

Learning Objectives

1. Describe the differences between artificial intelligence, machine learning, and deep learning
2. Provide examples of how machine learning can be used to solve problems (that traditional deterministic programming cannot)
3. Provide examples of how embedded machine learning can be used to solve problems (where other forms of machine learning would be limited or inappropriate)
4. Describe the limitations of machine learning
5. Describe the ethical concerns of machine learning
6. Describe the differences between supervised and unsupervised machine learning

Section 1: Machine Learning on the Edge

Lecture Material

ID	Description	Links	Attribution
1.1.1	What is machine learning	video slides	[1]
1.1.2	Machine learning on embedded devices	video slides	[1]
1.1.3	What is tiny machine learning	slides	[3]
1.1.4	Tinyml case studies	doc	[3]
1.1.5	How do we enable tinyml	slides	[3]

Exercises and Problems

ID	Description	Links	Attribution
1.1.7	Example assessment questions	doc	[3]

Section 2: Limitations and Ethics

Lecture Material

ID	Description	Links	Attribution
1.2.1	Limitations and ethics	video slides	[1]
1.2.2	What am I building?	slides	[3]
1.2.3	Who am I building this for?	slides	[3]
1.2.4	What are the consequences?	slides	[3]
1.2.5	The limitations of machine learning	blog
1.2.6	The future of AI; bias amplification and algorithm determinism	blog

Exercises and Problems

ID	Description	Links	Attribution
1.2.7	Example assessment questions	doc	[3]

Section 3: Getting Started with Colab

Lecture Material

ID	Description	Links	Attribution
1.3.1	Getting Started with Google Colab	video
1.3.2	Intro to colab	slides	[3]
1.3.3	Welcome to Colab!	colab
1.3.4	Colab tips	doc	[3]
1.3.5	Why TensorFlow?	video
1.3.6	Sample tensorflow code	doc	[3]

Exercises and Problems

ID	Description	Links	Attribution
1.3.7	101 exercises for Python fundamentals	colab

Module 2: Getting Started with Deep Learning

This module provides an overview of neural networks and how they can be used to make predictions. Simple examples are given in Python (Google Colab) for students to play with and learn from. If you do not wish to explore basic machine learning with Keras in Google Colab, you may skip this module to move on to using the Edge Impulse graphical environment. Note that some later exercises rely on Google Colab for curating data, visualizing neural networks, etc.

Learning Objectives

1. Provide examples of how machine learning can be used to solve problems (that traditional deterministic programming cannot)
2. Provide examples of how embedded machine learning can be used to solve problems (where other forms of machine learning would be limited or inappropriate)
3. Describe challenges associated with running machine learning algorithms on embedded systems
4. Describe broadly how a mathematical model can be used to generalize trends in data
5. Explain how the training process results from minimizing a loss function
6. Describe why datasets should be broken up into training, validation, and test sets
7. Explain how overfitting occurs and how to identify it
8. Demonstrate the ability to train a dense neural network using Keras and TensorFlow

Section 1: Machine Learning Paradigm

Lecture Material

ID	Description	Links	Attribution
2.1.1	The machine learning paradigm	slides	[3]
2.1.2	Finding patterns	doc	[3]
2.1.3	Thinking about loss	slides	[3]
2.1.4	Minimizing loss	slides	[3]
2.1.5	First neural network	slides	[3]
2.1.6	More neural networks	doc	[3]
2.1.7	Neural networks in action	doc	[3]

Exercises and Problems

ID	Description	Links	Attribution
2.1.8	Exploring loss	colab	[3]
2.1.9	Minimizing loss	colab	[3]
2.1.10	Linear regression	colab	[3]
2.1.11	Solution linear regression	doc	[3]
2.1.12	Example assessment questions	doc	[3]

Section 2: Building Blocks of Deep Learning

Lecture Material

ID	Description	Links	Attribution
2.2.1	Introduction to neural networks	slides	[1]
2.2.2	Initialization and learning	doc	[3]
2.2.3	Understanding neurons in code	slides	[3]
2.2.4	Neural network in code	doc	[3]
2.2.5	Introduction to classification	slides	[3]
2.2.6	Training validation and test data	slides	[3]
2.2.7	Realities of coding	doc	[3]

Exercises and Problems

ID	Description	Links	Attribution
2.2.8	Neurons in action	colab	[3]
2.2.9	Multi layer neural network	colab	[3]
2.2.10	Dense neural network	colab	[3]
2.2.11	Challenge: explore neural networks	colab	[3]
2.2.12	Solution: explore neural networks	doc	[3]
2.2.13	Example assessment questions	doc	[3]

Section 3: Embedded Machine Learning Challenges

Lecture Material

ID	Description	Links	Attribution
2.3.1	Challenges for tinyml a	slides	[3]
2.3.2	Challenges for tinyml b	slides	[3]
2.3.3	Challenges for tinyml c	slides	[3]
2.3.4	Challenges for tinyml d	slides	[3]

Exercises and Problems

ID	Description	Links	Attribution
2.3.5	Example assessment questions	doc	[3]

Module 3: Machine Learning Workflow

In this module, students will get an understanding of how data is collected and used to train a machine learning model. They will have the opportunity to collect their own dataset, upload it to Edge Impulse, and train a model using the graphical interface. From there, they will learn how to evaluate a model using a confusion matrix to calculate precision, recall, accuracy, and F1 scores.

Learning Objectives

1. Provide examples of how embedded machine learning can be used to solve problems (where other forms of machine learning would be limited or inappropriate)
2. Describe challenges associated with running machine learning algorithms on embedded systems
3. Describe why datasets should be broken up into training, validation, and test sets
4. Explain how overfitting occurs and how to identify it
5. Describe broadly what happens during machine learning model training
6. Describe the difference between model training and inference
7. Describe why test and validation datasets are needed
8. Evaluate a model's performance by calculating accuracy, precision, recall, and F1 scores
9. Demonstrate the ability to train a machine learning model with a given dataset and evaluate its performance

Section 1: Machine Learning Workflow

Lecture Material

ID	Description	Links	Attribution
3.1.1	Tinyml applications	slides	[3]
3.1.2	Role of sensors	doc	[3]
3.1.3	Machine learning lifecycle	slides	[3]
3.1.4	Machine learning lifecycle	doc	[3]
3.1.5	Machine learning workflow	doc	[3]

Exercises and Problems

ID	Description	Links	Attribution
3.1.6	Example assessment questions	doc	[3]

Section 2: Data Collection

Lecture Material

ID	Description	Links	Attribution
3.2.1	Introduction to data engineering	doc	[3]
3.2.2	What is data engineering	slides	[3]
3.2.3	Using existing datasets	slides	[3]
3.2.4	Responsible data collection	slides	[3]
3.2.5	Getting started with edge impulse	video slides	[1]
3.2.6	Data collection with edge impulse	video slides	[1]

Exercises and Problems

ID	Description	Links	Attribution
3.2.7	Example assessment questions	doc	[1]

Section 3: Model Training and Evaluation

Lecture Material

ID	Description	Links	Attribution
3.3.1	Feature extraction from motion data	video slides	[1]
3.3.2	Feature selection in Edge Impulse	video tutorial	[1]
3.3.3	Machine learning pipeline	video slides	[1]
3.3.4	Model training in edge impulse	video slides	[1]
3.3.5	How to evaluate a model	video slides	[1]
3.3.6	Underfitting and overfitting	video slides	[1]

Exercises and Problems

ID	Description	Links	Attribution
3.3.7	Project: Motion detection	doc	[1]
3.3.8	Example assessment questions	doc	[1]

Module 4: Model Deployment

This module covers why quantization is important for models running on embedded systems and some of the limitations. It also shows how to use a model for inference and set an appropriate threshold to minimize false positives or false negatives, depending on the system requirements. Finally, it covers the steps to deploy a model trained on Edge Impulse to an Arduino board.

Learning Objectives

1. Provide examples of how embedded machine learning can be used to solve problems (where other forms of machine learning would be limited or inappropriate)
2. Describe challenges associated with running machine learning algorithms on embedded systems
3. Describe broadly what happens during machine learning model training
4. Describe the difference between model training and inference
5. Demonstrate the ability to perform inference on an embedded system to solve a problem

Section 1: Quantization

Lecture Material

ID	Description	Links	Attribution
4.1.1	Why quantization	doc	[3]
4.1.2	Post-training quantization	slides	[3]
4.1.3	Quantization-aware training	slides	[3]
4.1.4	TensorFlow vs TensorFlow Lite	slides	[3]
4.1.5	TensorFlow computational graph	doc	[3]

Exercises and Problems

ID	Description	Links	Attribution
4.1.6	Post-training quantization	colab	[3]
4.1.7	Example assessment questions	doc	[3]

Section 2: Embedded Microcontrollers

Lecture Material

ID	Description	Links	Attribution
4.2.1	Embedded systems	slides	[3]
4.2.2	Diversity of embedded systems	doc	[3]
4.2.3	Embedded computing hardware	slides	[3]
4.2.4	Embedded microcontrollers	doc	[3]
4.2.5	TinyML kit peripherals	doc	[3]
4.2.6	TinyML kit peripherals	slides	[3]
4.2.7	Arduino core, frameworks, mbedOS, and bare metal	doc	[3]
4.2.8	Embedded ML software	slides	[3]

Exercises and Problems

ID	Description	Links	Attribution
4.2.8	Embedded ml software	slides	[3]
4.2.9	Example assessment questions	doc	[3]

Section 3: Deploying a Model to an Arduino Board

Lecture Material

ID	Description	Links	Attribution
4.3.1	Using a model for inference	video slides	[1]
4.3.2	Testing inference with a smartphone	video	[1]
4.3.3	Deploy model to arduino	video slides	[1]
4.3.4	Deploy model to Arduino	tutorial

Exercises and Problems

ID	Description	Links	Attribution
4.3.5	Example assessment questions	doc	[1]

Module 5: Anomaly Detection

This module describes several approaches to anomaly detection and why we might want to use it in embedded machine learning.

Learning Objectives

1. Provide examples of how embedded machine learning can be used to solve problems (where other forms of machine learning would be limited or inappropriate)
2. Describe challenges associated with running machine learning algorithms on embedded systems
3. Describe broadly what happens during machine learning model training
4. Describe the difference between model training and inference
5. Demonstrate the ability to perform inference on an embedded system to solve a problem
6. Describe how anomaly detection can be used to solve problems

Section 1: Introduction to Anomaly Detection

Lecture Material

ID	Description	Links	Attribution
5.1.1	Introduction to anomaly detection	doc	[3]
5.1.2	What is anomaly detection?	slides	[3]
5.1.3	Challenges with anomaly detection	slides	[3]
5.1.4	Industry and TinyML	doc	[3]
5.1.5	Anomaly detection datasets	slides	[3]
5.1.6	MIMII dataset paper	doc	[3]
5.1.7	Real and synthetic data	doc	[3]

Exercises and Problems

ID	Description	Links	Attribution
5.1.8	Example assessment questions	doc	[3]

Section 2: K-means Clustering and Autoencoders

Lecture Material

ID	Description	Links	Attribution
5.2.1	K-means clustering	slides
5.2.2	Autoencoders	slides	[3]
5.2.3	Autoencoder model architecture	doc	[3]

Exercises and Problems

ID	Description	Links	Attribution
5.2.4	K-means clustering for anomaly detection	colab	[3]
5.2.5	Autoencoders for anomaly detection	colab	[3]
5.2.6	Challenge autoencoders	colab	[3]
5.2.7	Solution autoencoders	doc	[3]
5.2.8	Example assessment questions	doc	[3]

Section 3: Anomaly Detection in Edge Impulse

Lecture Material

ID	Description	Links	Attribution
5.3.1	Anomaly detection in edge impulse	video slides	[1]
5.3.2	Industrial embedded machine learning demo	video	[1]

Exercises and Problems

ID	Description	Links	Attribution
5.3.3	Project: Motion classification and anomaly detection	doc	[1]

Module 6: Image Classification with Deep Learning

This module introduces the concept of image classification, why it is important in machine learning, and how it can be used to solve problems. Convolution and pooling operations are covered, which form the building blocks for convolutional neural networks (CNNs). Saliency maps and Grad-CAM are offered as two techniques for visualizing the inner gs of CNNs. Data augmentation is introduced as a method for generating new data from existing data to train a more robust model. Finally, transfer learning is shown as a way of reusing pretrained models.

Learning Objectives

1. Describe the differences between image classification, object detection, and image segmentation
2. Describe how embedded computer vision can be used to solve problems
3. Describe how convolution and pooling operations are used to filter and downsample images
4. Describe how convolutional neural networks differ from dense neural networks and how they can be used to solve computer vision problems

Section 1: Image Classification

Lecture Material

ID	Description	Links	Attribution
6.1.1	What is computer vision?	video slides	[2]
6.1.2	Overview of digital images	video slides	[2]
6.1.3	Dataset collection	video slides	[2]
6.1.4	Overview of image classification	video slides	[2]
6.1.5	Training an image classifier with Keras	video	[2]

Exercises and Problems

ID	Description	Links	Attribution
6.1.6	Example assessment questions	doc	[2]

Section 2: Convolutional Neural Network (CNN)

Lecture Material

ID	Description	Links	Attribution
6.2.1	Image convolution	video slides	[2]
6.2.2	Pooling layer	video slides	[2]
6.2.3	Convolutional neural network	video slides	[2]
6.2.4	CNN in keras	slides	[3]
6.2.5	Mapping features to labels	doc	[3]
6.2.6	Training a CNN in Edge Impulse	video doc	[2]

Exercises and Problems

ID	Description	Links	Attribution
6.2.7	Exploring convolutions	colab	[3]
6.2.8	Convolutional neural networks	colab	[3]
6.2.9	Challenge: CNN	colab	[3]
6.2.10	Solution: CNN	doc	[3]
6.2.11	Example assessment questions	doc	[2]

Section 3: Analyzing CNNs, Data Augmentation, and Transfer Learning

Lecture Material

ID	Description	Links	Attribution
6.3.1	CNN visualizations	video slides	[2]
6.3.2	Data augmentation	video slides	[2]
6.3.3	TensorFlow datasets	doc	[3]
6.3.4	Avoiding overfitting with data augmentation	slides	[3]
6.3.5	Dropout regularization	doc	[3]
6.3.6	Exploring loss functions and optimizers	doc	[3]
6.3.7	Transfer learning and MobileNet	video slides	[2]
6.3.8	Transfer learning with Edge Impulse	video slides	[2]

Exercises and Problems

ID	Description	Links	Attribution
6.3.9	Saliency and Grad-CAM	colab	[2]
6.3.10	Image transforms demo	colab	[2]
6.3.11	Challenge: image data augmentation	colab	[2]
6.3.12	Solution: image data augmentation	colab	[2]
6.3.13	Example assessment questions	doc	[2]

Module 7: Object Detection and Image Segmentation

In this module, we look at object detection, how it differs from image classification, and the unique set of problems it solves. We also briefly examine image segmentation and discuss constrained object detection. Finally, we look at responsible AI as it relates to computer vision and AI at large.

Learning Objectives

1. Describe the differences between image classification, object detection, and image segmentation
2. Describe how embedded computer vision can be used to solve problems
3. Describe how image segmentation can be used to solve problems
4. Describe how convolution and pooling operations are used to filter and downsample images
5. Describe how convolutional neural networks differ from dense neural networks and how they can be used to solve computer vision problems
6. Describe the limitations of machine learning
7. Describe the ethical concerns of machine learning
8. Describe the requirements for collecting a good dataset and what factors can create a biased dataset

Section 1: Introduction to Object Detection

Lecture Material

ID	Description	Links	Attribution
7.1.1	Introduction to object detection	video slides	[2]
7.1.2	Object detection performance metrics	video slides	[2]
7.1.3	Object detection models	video slides	[2]
7.1.4	Training an object detection model	video slides	[2]
7.1.5	Digging deeper into object detection	doc	[2]

Exercises and Problems

ID	Description	Links	Attribution
7.1.6	Example assessment questions	doc	[2]

Section 2: Image Segmentation and Constrained Object Detection

Lecture Material

ID	Description	Links	Attribution
7.2.1	Image segmentation	video slides	[2]
7.2.2	Multi-stage Inference Demo	video	[2]
7.2.3	Reusing Representations with Mat Kelcey	video	[2]
7.2.4	tinyML Talks: Constrained Object Detection on Microcontrollers with FOMO	video

Exercises and Problems

ID	Description	Links	Attribution
7.2.5	Project: Deploy an object detection model	doc	[2]
7.2.6	Example assessment questions	doc	[2]

Section 3: Responsible AI

Lecture Material

ID	Description	Links	Attribution
7.3.1	Dataset collection	slides	[3]
7.3.2	The many faces of bias	doc	[3]
7.3.3	Biased datasets	slides	[3]
7.3.4	Model fairness	slides	[3]

Exercises and Problems

ID	Description	Links	Attribution
7.3.5	Google what if tool	colab	[3]
7.3.6	Example assessment questions	doc	[3]

Module 8: Keyword Spotting

In this module, we create a functioning keyword spotting (also known as "wake word detection") system. To do so, we must introduce several concepts unique to audio digital signal processing and combine it with image classification techniques.

Learning Objectives

1. Describe how machine learning can be used to classify sounds
2. Describe how sound classification can be used to solve problems
3. Describe the major components in a keyword spotting system
4. Demonstrate the ability to train and deploy a sound classification system

Section 1: Audio Classification

Lecture Material

ID	Description	Links	Attribution
8.1.1	Introduction to audio classification	slides	[1]
8.1.2	Audio data capture	slides	[1]
8.1.3	What is keyword spotting	slides	[3]
8.1.4	Keyword spotting challenges	slides	[3]
8.1.5	Keyword spotting application architecture	doc	[3]
8.1.6	Keyword spotting datasets	slides	[3]
8.1.7	Keyword spotting dataset creation	doc	[3]

Exercises and Problems

ID	Description	Links	Attribution
8.1.8	Example assessment questions	doc	[1] [3]

Section 2: Spectrograms and MFCCs

Lecture Material

ID	Description	Links	Attribution
8.2.1	Keyword spotting data collection	slides	[3]
8.2.2	Spectrograms and mfccs	doc	[3]
8.2.3	Keyword spotting model	slides	[3]
8.2.4	Audio feature extraction	slides	[1]
8.2.5	Review of convolutional neural networks	slides	[1]
8.2.6	Modifying the neural network	slides	[1]

Exercises and Problems

ID	Description	Links	Attribution
8.2.7	Spectrograms and mfccs	colab	[3]
8.2.8	Example assessment questions	doc	[1] [3]

Section 3: Deploying a Keyword Spotting System

Lecture Material

ID	Description	Links	Attribution
8.3.1	Deploy audio classifier	slides	[1]
8.3.2	Implementation strategies	slides	[1]
8.3.3	Metrics for keyword spotting	slides	[3]
8.3.4	Streaming audio	slides	[3]
8.3.5	Cascade architectures	slides	[3]
8.3.6	Keyword spotting in the big picture	doc	[3]

Exercises and Problems

ID	Description	Links	Attribution
8.3.7	Project: Sound classification	doc	[1]
8.3.8	Example assessment questions	doc	[1] [3]

Attribution

[1] Slides and written material for "Introduction to Embedded Machine Learning" by Edge Impulse is licensed under CC BY-NC-SA 4.0

[2] Slides and written material for "Computer Vision with Embedded Machine Learning" by Edge Impulse is licensed under CC BY-NC-SA 4.0

[3] Slides and written material for "TinyML Courseware" by Prof. Vijay Janapa Reddi of Harvard University is licensed under CC BY-NC-SA 4.0