Hi Edge Impulse!
I sent a mail with same subject as this post on May 3 (to your e-mail “hello [at] edgeimpulse [dot] com”) but haven’t gotten any response yet, perhaps it went into your spam folder?
Would appreciate a response, either as e-mail or to this post. If you prefer, I can also post more or less my whole e-mail here in the forum.
Thx!
Thomas Vikström
Hi @noone,
I replied to you email yesterday at 3:56 PM (Paris time), can you make sure the answer did not arrive in your spam?
In any cases, I copy the response here as it may interest some other teachers:
Hello!
I am also a teacher in a University in Lille, northern France called Junia. I teach Embedded Programming and IoT.
Since last year I’ve also been giving an Introduction to Embedded ML laboratory using Edge Impulse.
To answer your questions:
Question 1: Can I use Edge Impulse in an educational environment?
Of course, each students can have their own accounts and they can create 5 project each.
How many projects per students you had in mind?
Question 2: What edge devices do you recommend?
For additional hardware, the Arduino Nano 33 BLE (https://store.arduino.cc/arduino-nano-33-ble) is probably the easiest one to use and comes with an accelerometer and a microphone. And of course you can plug the classic Arduino compatible sensors.
In my case, I have been using for my Embedded Programming classes the ESP8266 and ESP32-Cam but they are not officially supported by Edge Impulse yet (although it works using the compiled C++ library but they have limited resources).
Question 3: Any other recommendations?
Do you know if Edge Impulse has been used in education before? Any recommendations or pointers you can give?
Several teachers are currently using Edge Impulse with their students. I’d recommend you have a look at the Coursera Course (https://www.coursera.org/learn/introduction-to-embedded-machine-learning?) which is great. (I took the course in March and every steps is well explained).
You can also have a look at the material I used with my students: https://github.com/luisomoreau/Embedded-Machine-Learning-Lab
Let me know if you have any further questions.
Best regards,
Louis
Thx @louis, I did receive the e-mail and read it earlier today. Appreciate you took time for this!
I also think it’s good to continue communication here in the forum as it might be of interest to others.
Comments to Question 1: I found out this week that versioning is possible, this reduces the amount of projects needed. I believe the limits of 5 projects/student is ok, but as teacher I’d need more to be able to test and prepare different scenarios. I’m estimating that 15-20 for me as teacher would be enough.
Comments to Question 2: Price wise Arduino Nano 33 BLE is great, but as it does not have a camera, it feels a bit limited. I though suppose there are camera shields compatible with this Arduino. I personally have a Himax WE-I Plus which does have a camera, but the price is obviously a bit higher. Regardless of device, I’d want some type of enclosure as the device will be handled by many students in a classroom or lab setting.
Comments to Question 3: I’ll take a look at the Coursera Course, great tip! Many thanks for the github-link, your material looks awesome!
Can I use your material in my own teaching? Probably I’d in that case either create an own github-repository or more probable put in our Moodle.
Hello @noone,
Question 1: I’m estimating that 15-20 for me as teacher would be enough.
Let me loop internally with the team. This question opens a larger topic on our side around an education program. We will definitely talk about this soon but in the meantime, if you have some ideas of what features you would need, let me know.
In my case, I know this could be useful:
Question 2 : Price wise Arduino Nano 33 BLE is great, but as it does not have a camera
True, depending on your budget, you can get the Arduino Nano 33 BLE (around 18€) and an ESP32 CAM (around 5/10€).
I know many schools universities also own some Raspberry Pi which is great and supported by Edge Impulse (and because of the CPU, you can run some Object Detection models too). So if you already have some RPi, just get some Rapsberry Pi cameras.
Question 3: You can use extensively and without limitations any project that are in my Github repository, especially if it is for educational purposes 
Best regards,
Louis
Thx for your comments!
Related to educational needs (Q1), it is a bit too early for me to say, but at least the points you wrote about would be good, perhaps also slightly more computing resources than in regular free accounts.
Q2: Budget wise the Arduino devices you mention would most probably fit. We do have some Raspberrys as well (don’t remember the model), so getting a RPi cameras would be doable.
Once the curriculum for this course is clearer, I’d have a better view of the minimum requirements. I have probably not mentioned that the course would be “Fundamentals of AI - a hands-on approach” (working course title), so we would not go into all nitty-gritty details, but start simple and then build on top of that. This course I’m planning would also be part of my own research.
Q3 Thx for letting me use your material for educational purposes 
Kind regards,
Thomas
Hi @noone, I teach Machine Learning to a small after school group of High School students. Advertising plug: I might be teaching a 9am-noon PST class for teenagers the first week of July, solely based on Edge Impulse.
Some of it can be taught without a microcontroller as a laptop and cell phone is enough to get started. Also the final impulse can be put on a web page using Javascript and WASM. Then you can see what the students are producing.
A paid account allows multiple logins to share one impulse. That is how it is supposed to be done. Really interesting for all the students to train one impulse in real time. Not sure if Edge Impulse has an education pricing, but they would probably let you test an impulse.
Presently a few of my students have the Nano33Ble, but I have moved to the full Arduino Pro, including a class set of the PortentaH7, with LoRa Vision Shield. So I will be making a full Robotics and Machine learning curriculum around these micro-controllers. Hopefully ready for Jan 2022.
I used to teach this using TensorFlowJS, so Edge Impulse has been a major improvement in the students being able to make their own models.
Easiest to contact me using twitter @rocksetta, or here.
Thx @Rocksetta for your input!
Yeah, it would be great to be able to share an impulse between students, hopefully Edge Impulse can let me test this at some point (pinging @louis). I don’t have any budget yet (apart from my own thin wallet), but I believe my university would be able to purchase some type of educational license if needed.
I’m right now writing the plan for my research, and once that is approved I’ll start planning the course itself.
I’m anyhow thinking to start the first lecture by introducing the students to Lobe (by Microsoft) and use that to quickly differentiate between images from cats and dogs, persons with and without mask etc.
Soon after we’d probably move over to Edge Impulse and use mobile phones for data gathering. Might in the end be that the students would not use any other edge devices at all in this pilot course. Instead I’d as teacher demonstrate with the devices I have (OpenMV Cam + AMG8833 for thermal data, Himax + a Bangle smartwatch if I get it to work).