Week 7 - Machine Learning and Teachable Machine

This is the seventh week of Digital Practices.

This Week:

• Description:

We'll be using teachable machine to explore machine learning. It doesn’t require coding for the basic functionality. We also had a pre-workshop task which is to use p5.js.

• Reflection:

Using this shows how we can identify different things based on images. What is interesting is the Shiba Inu dog which it thinks is a cat. For this, adding more image samples would help it learn better and identify whether or not it is a cat or dog. Another example is with the fruits, since I only added red apples when I put in a photo of a red pear it thought that the pear was an apple. Increasing this to a sample size of 3 for example helped in this case. For the cat and dog, it is a lot more difficult since there are a lot of different types of cats and dogs. So we would need a bigger sample size to help the machine to learn. p5.js is pretty interesting to use, it uses the teachable machine model and the webcam to identify if someone is staring at the webcam or not. We can also make our own drawing using this and add this to the website.

• Work to be done for next week:

I could try looking at other areas of Teachable Machine, such as sound.

Overall Thoughts and Images:

I thought this was interesting, the more sample sizes there is the better the prediction is. So next time I could add in images of a Shiba Inu and the algorithm will recognise that it's a dog rather than a cat.

This is an image of the pre workshop task.

This is an image of the apple, orange and pear training model.

This is an image of the cat and dog training model.