Microsoft has recently released a free app, which allows anyone to train an ML model, without a single line of code.
The model can be transferred to various platforms and integrated into your application.
Lobe, acquired by Microsoft back in 2018, comes with a fairly simple interface: you name the project, feed it with photos and train the model to identify what it sees.
You can check the model you have trained with your camera or images you upload to it, and see how it classifies them in real time. If the fit is incorrect, you can correct it, and the model will improve its analysis.
I've tried to build a simple model and teach my little girl the different types of cutlery (spoon, fork, knife, etc.). After few back and forth and multiple photos, the model got approx. 99% accuracy, and it only took me 5 minutes with not a single line of code.
If you're concerned about privacy, Microsoft claims the software is working locally on your device and non of the data (/photos) is uploaded to the cloud.
Anyway, the possibilities are endless. I'm now "mapping" my hardware tools to help my kids identify which is which :) Give it a go yourselves...