[MakeIT] TinyML tutorial : How to forecast weather with Arduino nano 33 BLE Sense?

Hi everybody,

I am glad to share you the second video of MakeIT my YouTube channel.

Today we go deeper into tiny machine learning, the aim of this video is to forecast weather with 3 parameters as input.
In this video we talk about several topics such as :

  • Setting up the Kaggle API
  • Downloading datasets from kaggle
  • Data prep (cleaning, processing…)
  • Building neural network and train it
  • Deploying a machine learning model on nano33 BLE

The neural network creating and data downloading is done using Python, Tensorflow and Google colaboratory. Then we deploy the model on Arduino Nano 33 BLE Sense using Visual Studio code and PlatformIO.

Vijay Janapa Reddi’s TinyML courses from edX really helped me understanding the key of TinyML and what are the challenges of this fields. It is an awesome program with 3 courses I have attended ! I really encourage you to follow it !

Video here

And if you have never done any TinyML project I really encourage you to see my first video about deploying the simplest neural network on Nano33, this is a good way to start.

Next tutorial will be : “How to build your own datasets with Datasets Builder (my homemade software) ?” So stay tunned !


@BaptisteZloch, I watched your MakeIt channel video and now I am hooked. I look forward to your upcoming youtube on data collection. I am in the middle of trying to get an esp32 to post data in CBOR format to an edge impulse data ingestion service. (or to my server).

I am coming to the opinion that all of my IOT devices that are running my heating system , monitoring my greenhouse and watering my garden should be collecting and shipping out data to a server whenever something changes in the sensor inputs. Eventually there will be enough data to start making useful predictions on models yet to be devised.