Apple Adds 20 Core ML Models to Hugging Face, Expanding Open Source AI Contributions

Apple has bolstered its commitment to the open-source AI community by contributing 20 new Core Machine Learning (Core ML) models to Hugging Face. This significant update follows their earlier release of four Open Source Efficient LLMs (OpenELMs) in April 2024, which were also shared via Hugging Face.

What Changed and Why It Matters

The latest addition of these 20 models is a major step forward for Apple's engagement with the AI research community. These models are designed to enhance text and image processing capabilities, making them particularly useful for developers working on applications that require robust visual and linguistic understanding.

• Image Classification: Models that can identify and categorize objects within images.
• Semantic Segmentation: Tools that can segment images into distinct regions, each corresponding to a different object or part of the scene.
• Text Processing: Enhancements in natural language processing (NLP) for better text analysis and generation.

Key Details

• Image Classification Models:

  • These models are optimized for tasks like identifying objects, people, and scenes within images.
  • They can be used to develop applications that automatically tag photos or remove backgrounds.

• Semantic Segmentation Models:

  • These tools segment images into meaningful parts, such as separating the foreground from the background.
  • Useful in applications like augmented reality (AR) and photo editing software.

• Text Processing Models:

  • Enhanced NLP capabilities for tasks like sentiment analysis, language translation, and text summarization.
  • Can improve the performance of chatbots, virtual assistants, and content generation tools.

Practical Applications

Clement Delangue, co-founder and CEO of Hugging Face, highlighted the potential impact of these new models:

• Background Removal: Apps that can seamlessly remove unwanted backgrounds from photos.
• Object Identification: Tools that instantly identify objects in front of you and provide their names, even in foreign languages.

Context and Background

This release is part of Apple's ongoing efforts to contribute to the AI research community. In April 2024, Apple released four OpenELMs via Hugging Face, which are efficient large language models (LLMs) designed for various AI tasks. Additionally, in October 2023, Apple researchers published "Ferret," a large language model for image queries, on GitHub.

• Ferret: Initially released to handle image-based queries, Ferret has since been updated to improve its capabilities.
• Research Papers: Apple has also published several research papers on topics like generative AI animation tools and the creation of AI avatars. These advancements may have contributed to features in Apple Intelligence, such as the forthcoming Genmoji feature in iOS 18.

Future Implications

While this is Apple's first public release since announcing Apple Intelligence at WWDC, it underscores the company's commitment to advancing AI technology through open-source collaboration. The addition of these models to Hugging Face not only benefits the research community but also paves the way for more innovative and accessible AI applications.

Source