[PYTHON] Switch between multiple models with M5StickV

This article is the third day of M5Stack Advent Calendar 2019

Now, for the blind, there are various apps.

However, there seems to be no such thing as registering a PET bottle of tea that you want to drink and recognizing it. You can check what you have registered by using hardware such as Okamu My Eye 2, but the price is hundreds of thousands. It seems that the current situation is that it is expensive and difficult for students to use.

Therefore, available for about 3000 yen I created an image registration / recognition using M5StickV.

The model for recognition uses a WEB service called V-Training provided by M5Stack. V-Training creates a model that can discriminate up to 10 types of objects, so I created multiple models and made it possible to switch between them.

There are still some troublesome things such as V-Training itself and the process of making audio, but ・ Allows padding with a small number of images when recognizing flat objects such as logos (added) ・ Simplify the process of creating voice I hope to reduce the hassle.

The program and creation method are Open on GitHub.