Machine Learning in iOS
With Core ML, it's now possible to use machine learning in real-time on iOS devices. In this course, you'll learn to train models on a Mac and use them for data analysis. By Jessy Catterwaul.
Who is this for?
This course is for iOS developers who either have no experience with machine learning, or have used Core ML in some capacity, but have not trained their own models. You’ll need to be running macOS 10.14 Mojave or later in order to follow along. And you should be very comfortable using Swift and Xcode.
- Live image classification
- Training and tuning Core ML models
- Using custom models with Vision
- Augmenting data sets with Create ML
- Data curation for optimal training
- Creating Python environments for ML using Anaconda
- Using the Turi Create API in Jupyter Notebook
Part 1: Core ML and Create ML-
Welcome to the Machine Learning in iOS video course! In this introduction, find out what the course will cover.
Core ML Models
Use Core ML with a model that has already been trained for you, learning about how models are handled by Xcode.
Apple's Vision framework was introduced along with Core ML in iOS 11. They integrate easily for computer vision tasks.
Analyze the results of using a binary classifier. It will attempt to classify any image you show it, even if that results in nonsense!
Explore what happens when more than one object is in your image, or your model is able to recognize more than two classes.
Using other people’s models is often not sufficient. Create ML will allow you to train your own! Here you'll use it in a playground.
Collecting good data is key in ML. Hear about the curation process our team went through to provide the dataset for this course.
Reiterate the ways you've learned to work with ML models, whatever your budget for data collection and model training might be.
Part 2: Turi Create
If you’re going to get serious about Machine Learning, then you're going to need to be, or become, a Python coder! Turi Create has a Python API.
Get set up with a Python-based machine learning environment, based around Anaconda and Jupyter Notebook.
Despite the difference in programming languages (Python vs. Swift), Turi Create shares a lot with Create ML – including transfer learning.
Continue training your snacks classifier using Turi Create, in order to learn more about its API for managing models.
Confusion matrices are a really useful visualization of how well your model performs. Learn how to use them in a Jupyter notebook.
Turi Create's public API offers limited control over the model training process, but it's open source and hides some treasures!
You've gotten familiar with machine learning in two popular environments. But, armed with your Python knowledge, there's still a huge world of ML to explore!