Since the WWDC developer conference in early June, it has become clear: Apple has not only developed its own AI models and demonstratively published some of them online ahead of the presentation – the iPhone maker has also With Apple Intelligence There are even concrete plans to use its own AI model as soon as possible. Users with the US English language setting should be able to use it from the beginning of autumn. A week after WWDC, Apple has now uploaded 20 new CoreML models and four new datasets to an open source platform.
Advertisement
Apple chose the AI ​​open source platform Hugging Face for the upload. There The California-based company had already published four OpenELMs (Open Efficient Language Models) in April. Earlier, in October 2023, Apple uploaded the Ferrer AI model to GitHubWho can understand the user interface – a preview of a new feature announced for the voice assistant Siri.
These AI models include
Current new models include Depth Anything, which can calculate depth information in photos to distinguish between objects in the foreground and background. According to Apple, the model was trained with about 600,000 images that had labels attached. An additional 62 million training images were used without labels. ML models for classifying images FastVIT or DETR for semantic segmentation, that is, labeling or classification of an image at the pixel level, also go into the field of images. Data sets include FLAIR, which contains about 430,000 images from 51,000 Flickr users.

The newly published AI models can also be linked to functions previously presented by Apple, such as Genmojis, to create targeted emojis by entering text, with Image Wand, to generate images from text paragraphs or line sketches, and with Image Playground, to create images based on keywords, or Clean-up, a feature in the Photos app to remove unwanted things from images. What they all have in common is that they aim to process information on the device.
Unusually open
The fact that Apple is so open may be surprising to some who know the manufacturer primarily from its products – but this is not unusual considering other open source activities. Apple does the same This is the Swift programming language Published open source on GitHub and collaborated with the international community. Unlike product marketing, Apple is also open about future plans here.
In the case of AI models, Apple announced transparency at WWDC to increase trust in AI efforts. The publication on Hugging Face may also be aimed at encouraging developers to work with AI models in their apps and move data processing away from large cloud solutions to a more privacy-friendly way on the device. This development will also benefit Apple, as for the hardware manufacturer it may increase sales of existing devices that are equipped with the latest, more powerful neural engines for processing.
,Mk.I,