Apple, in collaboration with the University of California, Santa Barbara, has released an open-source artificial intelligence (AI) model called MLLM-Guided Image Editing, also known as “MGIE.” This model enables image editing similar to Photoshop using simple text commands.
Apple has been relatively reserved in the development of AI. Despite the hype surrounding ChatGPT last year, the company did not announce any major plans related to this technology. However, according to sources, Apple has its own chatbot named “Apple GPT,” inspired by ChatGPT. Furthermore, Tim Cook has announced that the company will soon unveil significant events related to AI.
It is yet unknown whether one of these events will involve an image editing tool using AI. However, the released MGIE model clearly indicates that Apple is conducting research and development in this area. Currently available AI tools often struggle with interpreting short commands, resulting in unsatisfactory results. MGIE is an innovative solution that utilizes multimodal language models (MLLMs) to understand text commands and training image data. As a result, MGIE can recognize natural commands, disregarding excessive descriptions.
The presented research examples demonstrate that MGIE can independently deduce instructions based on an input image. For example, given an image of pepperoni pizza and the instruction “make it healthier,” MGIE can infer that “it” refers to pepperoni pizza and “healthier” means adding vegetables. The result is an image of pepperoni pizza with green vegetables arranged on top.
In another comparative example, where the image depicts a forest edge and a calm body of water, other models do not produce the effect of a lightning reflection as instructed to “add lightning and make the water reflect the lightning.” However, MGIE successfully captures this effect.
The MGIE model is available as an open-source model on the GitHub platform and a demo version on the Hugging Face platform. This allows anyone to start experimenting with image editing using AI.
Apple continues its research and development in the field of AI, which may bring many innovative tools and applications in the future.
FAQ
The source of the article is from the blog combopop.com.br