Drag SqueezeNet. Xcode auto-generates a file for the model that includes classes for the input, output and main class. The main class includes various methods for making predictions. Go to CreateQuoteViewController. In this example, the classifier has a Find classificationRequest and replace the print statement with the code below to log the results:. Build and run the app and go through the steps to select a photo.
The console should log the top results:. Add the following to the CreateQuoteViewController extension:. The method runs on the main queue to ensure that the quote display update happens on the UI thread. Finally, call this method from classificationRequest and change request to the following:. Build and run the app. Select a photo with a lemon or lemon tree in it. If necessary, download one from the browser.
You should see the lemon quote selected instead of a random quote:. With Core ML 3, you can fine-tune an updatable model on the device during runtime. This means you can personalize the experience for each user. On-device personalization is the idea behind Face ID. Apple can ship a model down to the device that recognizes generic faces.
During Face ID set up, each user can fine-tune the model to recognize their face. This underscores the advantage of the privacy that on-device personalization brings. The classifier recognizes new drawings based on k-Nearest Neighbors, or k-NN. It does this by comparing feature vectors. Comparing the distance between feature vectors is a simple way to see if two objects are similar. The example below shows a spread of drawings classified as squares and circles.
Vibes uses an updatable drawing classifier with:. In Vibes, the user can add a shortcut by selecting an emoji then drawing three examples. Open AddStickerViewController. Next, open AddShortcutViewController. Finally, open CreateQuoteViewController. This removes the code that allows the user to move stickers around. Different models are good at different tasks.
For general object recognition, I recommend starting with MobileNet as its optimized for mobile devices.
To use one of these models, simply download it and drag and drop it into your Xcode project. You can then use Core ML to instantiate that model. To use it with vision we'll create a request just like in the Vision example, but this time it's a Core ML request with our model. We can then perform this request just like in the above example. As you can see, you can add object recognition to your app in just a few lines of code! Create ML is optimized for macOS and is integrated with Xcode's playgrounds, so it's the easiest way for iOS developers to train their own machine learning models.
To create an custom image classifier model, simply launch a new macOS playground and write these two lines of code:. A new pane shows up in the assistant editor where you can drag and drop images organized by class.
Each class needs to have its own folder. Drag and drop those folders and the training will start immediately. You can also tweak settings like how many iterations it will take, or whether it should create new input images by morphing the original ones to make the model more robust.
This is where transfer learning kicks in: Apple has their own network to recognize images, so Create ML is really fast and produces tiny models, because it leans heavily on the knowledge it already has. Once you have a model you can use Core ML to add it to your app like in the example above. Used for the same purpose as Create ML, but it's a little more flexible and supports more types of models. The catch is that it's a Python library, so no Swift support!
But don't worry if you never used Python. Setting up Turi Create couldn't be easier, and Python is a very simple language that you can get the hang of pretty fast. Turi Create is out of scope for this article, but it's good to know that you can get even more flexibility when you run into issues with Create ML.
As you can see, Apple offers a lot of tools with varying levels of flexibility. If you need general object detection, you can easily use Vision. If you want to detect a specific object, you can build your own model with Create ML and use Core ML in combination with Vision to detect that object. Trust me, it sounds a lot more complicated than it really is. Bring on-device machine learning features, like object detection in images and video, language analysis, and sound classification, to your app with just a few lines of code.
The easy-to-use app interface and models available for training make the process easier than ever, so all you need to get started is your training data. You can even take control of the training process with features like snapshots and previewing to help you visualize model training and accuracy.
Learn more. Models Get started with models from the research community that have been converted to Core ML. Browse models. View Code. Original Source MobileNetV2. Original Source SqueezeNet.
0コメント