Introduction to TinyML - Solomon Muhunyo Githu
1 like•1,212 views
The document discusses tinyML, a branch of machine learning focusing on running models on low-power microcontrollers for IoT applications. It highlights the benefits of tinyML, including reduced bandwidth, improved privacy, and energy efficiency, while listing various microcontroller units and frameworks such as TensorFlow Lite for microcontrollers. Additionally, it outlines the machine learning tasks applicable in this context and provides a project flow for deploying models on embedded devices.
Introduction to TinyML - Solomon Muhunyo Githu
- 1. Let's talk about TinyML Intelligence on Microcontrollers Hello.. Solomon Muhunyo Githu Researcher Waziup e.V. Contributor, Edge Impulse Expert Network June 2023.
- 2. Devices Users Simply, IoT is connecting devices to the internet so their data can be seen from anywhere in the world. These devices can also be controlled from anywhere through the internet. IoT architecture Cloud Process data and classify if a car or elephant is detected Car and elephant have been detected Car and elephant have been detected
- 3. Latency Energy saving By running ML models on the embedded devices we can get benefits such as: Solving IoT challenges Reduced bandwidth Data does not need to be transferred to the cloud Privacy Little or no internet connection may be required Microcontrollers require less power to perform Data is not transferred or stored on the cloud
- 4. Small devices Resource constrained devices TinyML is a branch of machine learning and embedded systems research that looks into the types of models that can be run on small, low-power devices like microcontrollers TinyML, What is it? Low-power devices
- 5. Personal assistant like Siri Industrial predictive maintenance Wildlife tracking Detecting crop diseases Healthcare Ocean life conservation Applications of TinyML The Future of ML is tiny and bright
- 6. Some cool TinyML MCUs Arduino Nano 33 BLE Sense Arduino Nicla Vision Arduino Nicla Sense ME Arduino Portenta H7 + Vision Shield Espressif ESP32 Espressif ESP-EYE Himax WE-I Plus Open MV Cam H7 Plus SiLabs xG24 Dev Kit Seeed Grove - Vision AI Module Sony's Spresense Syntiant Tiny ML Board Raspberry Pi Pico ......
- 7. So tiny, how? TF Lite for Microcontrollers is a modified version of the TensorFlow Lite framework that is meant to run on embedded devices with only a few tens of kilobytes of memory. It supports Android, IOS, Arduino etc.. Not only Python you can use C, C++ and JAVA Pretrained models TensorFlow is an open source framework developed by Google researchers to run machine learning, deep learning and other statistical and predictive analytics workloads. Credits: https://www.infoworld.com/article/3278008/what-is-tensorflow-the-machine-learning-library-explained.html
- 8. Credits: https://www.slideshare.net/janjongboom/adding-intelligence-to-your-lorawan-devices-the-things-conference-on-tour Machine Learning tasks Classification - what's happening right now? Anomaly detection - is this behavior out of the ordinary? 01 02 Forecasting - what will happen in the future? 03 Right algorithm Neural network - algorithm mimicking human brain K means clustering - tries to group similar kinds of items in form of clusters Regression - statistical technique that relates a dependent variable to one or more independent variables
- 9. Optimize model Reduce size Improve performance TinyML project flow Prepare data Train model (80% data) Validate model (20% data) Deploy model Inference at the Edge (MCU) .h5 or .pb Tensorflow model Convert model TensorFlow Lite converter Tests fails Tests pass With AutoML tools, it can't get simpler
- 10. Take a free course on introduction to embedded Machine Learning (link here) Next steps Scan me!
- 11. Thank You Solomon Muhunyo Githu