Hi everyone! Is there any learning resources with this tinyml kit? Will they be adding more edx course about tinyml using this kit? I found that the resources in tinyml field are quite relatively constrained, as I just found some like Pete’s ‘Tinyml’ book( the book is not completely compatible with the kit though) and the edx course only. How can we continue our learning with this kit? Are there any ways to learn about tinyml and Arduino with this kit rather than just putting the kit in the storeroom after this edx course?
Ps: I also found that the ov7675 is not so ‘popular’ in the Arduino projects, then why we are using this in the kit?
1 Like
Hi @Albert,
Great to have you here and thanks for asking the tough questions 
Yes, March 2nd we will be launching Course 3 and in this course we will be using the kit to train and deploy models onto the device:
Yeap, can’t agree more. Agree that the resources are quite limited. Have you checked out Education - TinyML Community where we have a few resources listed? Please feel to contribute if you find more.
Agreed that @petewarden’s book doesn’t cover this camera module, but rest assured everything we release on March 2nd is going to be replicating exactly what the exercises do in the book. In fact, we push the examples further by introducing new concepts like Multi-tenancy (running multiple models/sensors together), etc.
Great question!
We will be announcing a few projects that will build off of the kit as a starting point. These projects will be structured so that you can build on the knowledge you get from the “Deploying TinyML” course. So stay tuned for that.
Moreover, feel free to check out some of the projects that people are doing. Some starter samples are here:
Yet another great question!
The answer is $imple The OV camera module costs just $2.00 to $4.00, which helped us keep the kit cost really low. If we went with the ArduCam that is used in the TinyML book, then the BOM cost goes up, so that’s the reason. But again, everything you learn and gain from the course is still directly applicable. Nothing you the device itself changes what you learn in terms of model training, evaluation and deployment.
If others have thoughts, please feel free to chime.
Loads of good questions… have I answered them all?
Thanks again @Albert as your questions are bound to be on other peoples’ minds as well.
3 Likes
First of all, thanks @vjreddi for the long and detailed reply!
In fact, I have completed the course 1 and 2. I just received the kit yesterday 
and I am looking forward to the course 3!
Now I have learned more about the course content and other projects. I am glad that the difference between camera modules won’t be a big problem 
Lastly, thanks again @vjreddi for keeping the cost of the kit affordable, and most importantly, providing high-quality courses about tinyml so that everyone can learn about this emerging technology.
ps: is ov7675 and ov7670 interchangeable?
1 Like
I encourage you to check out what Evan (on of the HarvardX staff members) put out about how you can get the BOM directly: TinyML Hardware Kit - #38 by jesmith. It talks about how to get the individual parts if needed. One reason we put this together is because the cost can be cheaper when you cut out all the headaches with shipping and custom fees etc. All of which was well out of our control 
Anyway hope it helps folks looking into getting parts for the long run. I strongly encourage folks to get the kit just cause we have tested everything using it.
Yes, and no. You have to change the code a little bit to get it working. We in fact provide instructions on this as well in Course 3 just in case people aren’t able to get the OV7675 part… but again, prefer if people use the kit just cause we know it all works and we have shipped an Arduino library as well that has the necessary tests. etc.
Cheers!
I am looking forward to all the projects that will hopefully come out of doing Course 3!
2 Likes
While waiting for the course 3 (Deploying TinyML) on edx to start on March 2, I am taking Edge Impulse’s “Introduction to Embedded Machine Learning” in Coursera. This course is using Arduino Nano 33 BLE Sense. That’s awesome because edx course 3’s TinyML kit also includes the same Arduino board!
As a matter of fact I bought edx TinyML kit for course 3 and got the Arduino Nano 33 BLE Sense in the kit. However, since there’s a delay now in course 3, this Edge Impulse course came out in a good time! It’s giving me the opportunity to get started with TinyML deployment! It’s a great course in its own rightght. I became a huge fan of the Edge Impulse tool for TinyML!
Surely I can start the edx course 3 being more prepared now!
2 Likes
Thanks for sharing! The course seems fun!
@vjreddi I found some lectures videos on this website:
It is suitable for us as Edx learners to watch the videos as extra learning?
3 Likes
True. But the EdgeImpulse course practice exercises require their bootloader to be loaded on to the Arduino Nano 33 BLE Sense replacing the existing mbed os / bootloader.
I am not sure if this will have any implications with deploying TinyML models onto it in Course-3 of EdX. My guess at this point is that, it may need to be re-programed (flashed) with stock mbed os/bootloader, for it to function as desired in Course-3.
EdgeImpulse is an interactive visual development front end tool to do all the work we have done in courses one and two using Google Colab notebooks. The edgeimpulse studio simplifies Data collection, Classification, Model training, Testing and Deployment easier without needing to know what happens under the hood.
The course on Coursera also recommends the same kit (Arduino Tiny Machine Learning Kit), and offers a discount code to order the kit direct from Arduino store.
3 Likes
Edge Impulse is indeed a great tool.
In fact, in Course 4 (to be announced), which focuses on Advanced Topics in TinyML, we will touch on Edge Impulse as a way to improve MLOps… cause whatever we are going through is what is under the hood in Edge Impulse, and once you have that knowledge under the belt, then it is best to focus on high level tools like Edge Impulse as they allow you to focus on the task at hand without getting bogged in details.
4 Likes
Hi @Albert,
Indeed, that’s from the last iteration of the TinyML course we taught at Harvard. The lectures are great, and many of the concepts we cover will also be reflected in there. So consider the material to be quite complementary.
2 Likes
Glad to hear a course 4 may be in the works, I’ll start saving up now! 
2 Likes
Just audit it for free 
( You won’t really miss out anything. )
5 Likes