AI and Machine Learning for On-Device Development

Chapter 13. Create ML and Core ML for Simple iOS Apps

In this book so far, you’ve been looking at technologies that bring machine learning to multiple devices, so that you could use a single API to reach Android, iOS, embedded systems, microcontrollers, and more. This was made possible by the TensorFlow ecosystem, and in particular TensorFlow Lite, which underpins ML Kit, which you used as a higher level API. And while we didn’t go into embedded systems and microcontrollers, the concepts are the same, other than hardware limitations the smaller you go. To learn more about that space, check out the great book TinyML by Pete Warden and Daniel Situnayake (O’Reilly).

But I would be remiss if I didn’t at least cover the iOS-specific Create ML tool and Core ML libraries from Apple, which are designed to let you use ML models when creating apps for iOS, iPadOS or MacOS. Create ML in particular is a really nice visual tool that lets you create models without any prior ML programming experience.

We’ll look at a few scenarios, starting with creating a model that recognizes flowers, similar to the ones we did earlier with TensorFlow and TensorFlow Lite.

A Core ML Image Classifier Built Using Create ML

We’ll start with creating our model. We can do this codelessly using the Create ML tool. You can find this by right-clicking on Xcode in the doc, and then in the Open Developer Tool menu, you can find Create ML. See Figure 13-1.

Figure 13-1. Launching Create ML

When the tool launches, you’ll ...