AI and Machine Learning for On-Device Development

Chapter 12. Productizing Your App Using Firebase

So far in this book, you’ve explored using machine learning to create models, and you’ve looked into how to integrate them into Android or iOS apps using a variety of technologies. You could go low level with TensorFlow Lite, using the model directly, and dealing with the process of data conversion to and from the model. Or, for a number of common scenarios, you could take advantage of ML Kit to use a high-level API with an asynchronous programming methodology to make responsive applications easier to build. In all of these cases, though, you just built a very simple app that did inference in a single activity or view.

When it comes to productizing an app, you, of course, have to go much further, and Firebase is designed to be a cross-platform solution that intends to help you build, grow, and earn from your app.

And while a full discussion of Firebase is beyond the scope of this book, there is an important feature in Firebase that’s available in the free (aka Spark) tier that you can really take advantage of: custom model hosting.

Why Use Firebase Custom Model Hosting?

As you’ve seen throughout this book, creating an ML model to solve a problem for your users doesn’t have to be difficult. It’s relatively straightforward, thanks to tools like TensorFlow or TensorFlow Lite Model Maker that quickly train a model based on your data. What’s hard to do is to create the right model, on the obvious assumption that to be able to do this, ...