Image classification is fun! but hard...

The openMV Cam

This camera is pretty good for machine learning purposes. this openMv cam comes with an IDE that utilises python to program the machine learning models.

This is a really fun IDE to work in! Enough examples that helps you on your way to make your own projects. It even has an example that shows how you can connect an Arduino with I2C communication. That was a big win for me. Because I needed such a communication method!

I learned a lot in this phase:

• I learned some basic python code (syntax)
• I learned how to upload data to edge impulse and make/train models.
• I learned how to sent data to my Arduino with the python syntax.
• I learned how to utilise the results of my models that can make an effective example with a traffic light situation.
• I learned how to connect multiple devices to an I2C bus.

Starting off with little prototypes

Face detection with tensorflow that can move sensors in a certain direction:

This was a fun concept but did not utilise much machine learning. It's only face detection which is not an impressive machine learning model. So I started looking for a demo that could illustrate this image classification at it's best!

A traffic light system, object avoidance is the answer!

I made a traffic light using my second Arduino Uno. I made a dataset that had 3 outputs: green, orange, red

Final result

I had a working model that would drive forward if it is on the green light, stop if it's on an orange or red light.

Next up I'm going to train the model that it will avoid the Devine mug