AI and Machine Learning for On-Device Development

Chapter 9. Creating Custom Models

In earlier chapters you saw how to use turnkey models for image labeling, object detection, entity extraction, and more. What you didn’t see was how you might be able to use models that you had created yourself, and indeed how you might be able to create them by yourself. In this chapter, we’ll look at three scenarios for creating models, and then in Chapters 10 and 11, we’ll look at incorporating those models into Android or iOS apps.

Creating models from scratch can be difficult and very time consuming. It’s also the realm of pure TensorFlow development and is covered in lots of other books such as my book AI and Machine Learning for Coders (O’Reilly). If you aren’t creating from scratch, and in particular if you’re focused on mobile apps, there are some tools to assist you, and we’ll look at three of them in this chapter:

• TensorFlow Lite Model Maker is the preferred choice if you are building an app that fits a scenario that Model Maker supports. It’s not a generic tool for building any type of model, but is designed to support common use cases such as image classification, object detection, and more. It involves little to no neural-network-specific coding, and as such is a great place to start if you don’t want to learn that stuff yet!

• Creating models using Cloud AutoML, and in particular the tools in Cloud AutoML designed to minimize the amount of code you have to write and maintain. Similar to TensorFlow Model Maker, the scenarios here ...