Apple today introduced upgrades for the Core ML machine learning framework, including model encryption using Xcode and Core ML Model Deployment, a way to store and launch models and update AI independent of the app update cycle. AI within apps can power a range of features from classification of natural language or images to analysis of speech, sounds, and other media.
“In the past, you would have to push more app updates just to get the newer models in your user’s hands. Now with model deployment, you can quickly and easily update your models without updating the app itself,” Apple engineer Anil Katti said in a WWDC session.
Core ML Model Deployment also gives developers a way to group models into collections and offers targeted deployment for machine learning customized for operating system, device, region, app version, and other variables. Developers must enable Core ML Model Deployment in the Core ML API.
Also new in machine learning for Apple developers using the Create ML Mac app for training AI are templates for action classification for videos and style transfer for images and video. The action classification template uses pose estimation to track movement of the human body and can power apps for exercise or dance training. Developers can train a Create ML action classification template using videos with one human subject and one action type at a time.