Arduino Machine Learning Tutorial: Introduction to TinyML with Wio Terminal
Вставка
- Опубліковано 6 сер 2024
- Welcome to the Arduino Machine Learning Course with Wio Terminal! In this introductory video, we dive into the exciting world of Machine Learning on Microcontrollers with Arduino IDE. Join us as we explore the capabilities of the Wio Terminal and learn how to harness its power for gesture recognition using a light sensor.
🔗 Link to the blog article: www.hackster.io/dmitrywat/tin...
🔗 Artificial Nose Project: wiki.seeedstudio.com/Wio-Term...
🔗 Wio Terminal Specs: wiki.seeedstudio.com/Wio-Term...
Timestamps:
0:00 Intro
1:20 Why TinyML?
4:20 About Wio Terminal
5:10 Software overview
6:15 Preparing the environment
11:50 Gathering the data with data forwarder
14:16 Data processing blocks
19:16 Learning blocks
22:20 Deployment to device
25:20 Testing on device
25:36 Conclusion
Music in the video:
Future Garage - Nu-Garage, Post-Garage, Space Garage
• Music for Work - Artif...
Oscuro - Time Stood Still
Almost Vanished - Disclosure
Don't miss out on this insightful course! Subscribe to my channel for more exciting content on Wio Terminal, TinyML, and Arduino IDE. Leave your comments and questions below. - Наука та технологія
Hi there, this is Arijit, Ambassador from Edge Impulse. Nice explanation! Glad you liked using Edge Impulse. Looking forward for more. Keep it up! 👏👍
Hi, Arijit! Thank you very much for kind words! Already published second video, more coming.
Very good 👍
Thanks for the visit
Nice Explanation,
I want to get into TinyML, what are the languages that should be concentrate on?
so, you are showing the on the edge impulse, which is a great visual front-end tool for tinyML
but, you got to show what under the hood happening.
I hope you got my point.
Python for data processing and model training, C/C++ for inference on the device.
From an engineer standpoint, when creating an application/solution, the exact mechanics of how the inference is performed on the device might not be that important - unless you think you want to further optimize the performance.
But no worries, as it is mentioned in the video there are going to be 2-3 tutorials using pure Tensorflow Lite for Microcontrollers, where you get to see all the "under the hood" stuff.
@@HardwareaiThanks for the response, Great, waiting for the next videos
Will this also work on the recently announced Rasbarry pi Pico board?
Great question! I will try next week - there is portable C++ library as an option for on-device deployment in Edge Impulse, so that should work. And for Tensorflow Lite for microcontrollers, this take a look at this repo!
github.com/raspberrypi/pico-tflmicro
Hi
Arduino based on avr right?
Can we use tinyml in avr like atmega8 ? Why we dont that?!
Some Arduino boards are AVR based, while newer ones have mostly ARM cores. It is possible to run very simple traditional ML algorithms on AVR boards too, but it won't work for NN based models - too little RAM, FLASH.
@@Hardwareai ok
But in this model arduino uno used that based on avr and have same ram and flash
Is it possile to run this model on avr ?
You know i want to know how its possible ! Do we need a specific library or a specific method?
Hi, I have not been able to make the project work on my WIO, I get the following error:The size of your 'features' array is not correct. Expected 100 items, but had 0, I do not know how to currently implement the part of the light sensor to store in the features array, the example code don't work, can you recommend me something?
Hmmm. Example code from where did you use?
@@Hardwareai hi, this one: wiki.seeedstudio.com/Wio-Terminal-TinyML-EI-1/ but i think the raw_feature_get_data funtion are outdated
@@johncaipa Yeah, it's not being updated.. Try this link www.hackster.io/dmitrywat/tinyml-course-1-gesture-recognition-with-light-sensor-25cebb
Тебя смотрят много кто делает только первые шаги , зачем загружать уже имеющиеся . Показывай нормально.
Привет! Немного не понял твой коммент. + Нет, я думаю мой канал - не для делающих первые шаги. Первые шаги - это установка IDE, LED блинк и так далее. Таких видео и туториалов и без меня в сети полно)
Please demonstrate as fast as you can because I was not lost early enough in the video. Needless to say I stopped watching at the 10-minute mark.
Hi there! Yes, you're right on that one. In fact I've been researching some tips on how to improve audience engagement and that one is definitely on top of my list - demo the results earlier, so people would be more interested :)
@@Hardwareai My thoughts were on the pace of the presentation. For every keystroke/mouse click you rush through in the interest of time, the greater the percentage of people that will lose interest and the higher the likelihood of page closure. If necessary, create a Part I and Part II but don't rush them so they can follow along. I wrote and taught advanced technology classes for 23 years so consider this opinion well-founded. Good luck.