From the course: Machine Learning for iOS: Core ML and Create ML

Introduction to Core ML - iOS Tutorial

From the course: Machine Learning for iOS: Core ML and Create ML

Start my 1-month free trial

Introduction to Core ML

“

- [Instructor] Core ML is an SDK within Apple's development offerings that can be used, and is supported across iOS, macOS, tvOS, and watchOS. In short, the SDKs allow you to do machine learning in your applications built with Xcode. The way it works is that you feed the inputs into training your model, and the ML algorithm will generate outputs which are presented into an easy to understand interface. But what are all these different terms? Inputs, model, algorithm, and output, let's take a quick look. First you have your inputs, or the data, every device you use to generate data can be used for machine learning purposes. For example, every search you do on your device, or you moves you make on a game, these are all inputs. As of the latest version of the SDK, Core ML supports vision for analyzing images, sounds for identifying sounds, or converting them to text. Activity like gyroscope information, text and tabular, which is tables, information. In other words, you can apply machine learning to these inputs. Then you have a model, the model is how we apply machine learning algorithms to your inputs. You use a model to make predictions based on your input data. For example you can train a model to detect specific objects within a photo directly from its pixels. You can also convert models from other libraries into a Core ML model. And if you've built models outside of Apple's SDK's, take a look at the document on how to covert them. Finally, from these models you'll get outputs, or the result of the algorithms of the model. And this can be applied to hundreds of situations, or needs based on the available inputs. Or as it is often called, training sets of data. In our case, as of this version of the SDK, you can use the previously mentioned types of data. With Create ML you can test and train your model interactively, outside of the application environment. So even before integrating machine learning into an application, you can train your models and see the results, and then decide when you'd like to integrate it into your application. Again, if some or most of this doesn't make sense, it might become clearer as we work through our examples.

Contents