Model to Monitor toxic bloom water sources

I remember reading about the 300 elephants found dead. Their tusks were intact and there was no evidence of poison. The mystery was solved and the water source containing toxic bloom of cyanobacterium.

My project has two software components. Edge Pulse Machine Learning Model and Balena to track environmental variables i.e Temperature, Humidity and Pressure. The hardware used is the Raspberry Pi4 and the Raspberry Pi Sense Hat.

Creating a Model

1. Connect a Device

I was able to connect my Android phone by scanning a QR code with my phone. The project updates when I make changes on my laptop.

2. Data Acquisition

I decided to use images or datasets to train the model. I checked to make sure there was no copyright or permission required to use the images from Unsplash. I created two folders on my Windows 10 laptop called Training and Testing. This would be for uploading to Impulse Edge.

Uploading is simple. You just want to navigate to the folders and select the images to upload to the cloud.

Create Impulse

Image

Image

Transfer learning

I set the minimum confidence rate at 80%. If the confidence rate is below 80% the image will be tagged "uncertain"

Live Classification

Climate Change

There is a question of Climate Change causing water contaminated by toxic blue-green algae. So I decided to include Balena open source and cloud application to measure Temperature, Humidity and Pressure.

The image can be burned to the Raspberry Pi and can be found here: https://github.com/balenalabs/balena-sense

Burn the image to a SD card. I love using Etcher for this. Place the SD card and add power.

Create and deploy the application

Add a device and download the balenaOS disk image from the dashboard

This process creates a customized image configured for your application and device type and includes your network settings if you specified them.

Flash your SD card with the balenaOS disk image and boot the device

In the device details you should see a button:

This will bring you to a dashboard as below. The nice thing is you can use the url to access the dashboard any where in the world.

+