Screen recording at youtu.be/cpCeaR9KTF8
MobileNet is a light-weight machine learning model for performing image classification.
In this Machine Learning for Kids project, students can try MobileNet for themselves using the familiar educational low-code programming language Scratch.
A test box can be moved around using the mouse, and resized and stretched using arrow keys. When the mouse is clicked, the contents of the test box is submitted to the MobileNet model, and the top result is displayed. (The screen recording above is an example of what this looks like.)
By experimenting with their own images, students get first-hand experience of the sort of behaviour of this widely-used machine learning model, and an insight into the kinds of things it is good and bad at recognizing.
They’re also introduced to the imagenet set of images that was used to train the model, and the way that this was collected and labelled.
There are step-by-step instructions in the free “MobileNet” PDF available for download from machinelearningforkids.co.uk/worksheets. (For teachers who need to modify or customize the project for their own students, the original MS Word document is also available).
Please give it a try, and let me know what you think!
If you want a more light-hearted and fun variation of this project, the “I-Spy” project (also available for download) gets students to use the MobileNet model to make an I-Spy game. Or there are many other projects where students collect their own training images to train their own custom machine learning models.
Tags: mlforkids-tech