苹果和加州大学圣巴巴拉分校（UCSB）的研究人员联合开发并开源了一款名为MGIE（MLLM-Guided Image Editing）的图片编辑框架。这一框架的研发旨在解决指令引导不足的问题，通过将多模态大模型MLLM（Multimodal Large Language Model）应用于图像编辑领域。
据悉，MLLM模型通过学习获取简明的表达指令，并为图像编辑提供明确的视觉相关引导。通过端到端训练，这一模型能够同步更新并利用预期目标的潜在想象力来执行图像编辑。在人类指令的引导下，MGIE能够进行类似于Photoshop风格的修改、全局照片优化以及局部对象的修改，为用户提供更加灵活和高效的图像编辑工具。
这一开源的图片编辑框架引起了业界的广泛关注。专家表示，MGIE的推出将极大地促进图像编辑技术的发展，为用户提供更加智能和便捷的编辑方式。苹果和UCSB的合作也再次彰显了跨界合作的重要性，将不同领域的专业知识融合在一起，推动科技创新不断向前发展。
在未来，随着MGIE框架的不断优化和普及，相信图像编辑领域将迎来更多创新和突破，为用户带来更加丰富多彩的编辑体验。这一新闻也再次证明了科技的不断进步，将不断为人们的生活带来便利和乐趣。
英语如下：
News Title: Apple and UCSB Open Source MGIE Image Editing Framework, Revolutionizing Image Processing Technology!
Keywords: MGIE framework, multimodal large model, image editing
News Content: Apple and researchers from the University of California, Santa Barbara (UCSB) have jointly developed and open-sourced an image editing framework named MGIE (MLLM-Guided Image Editing). The development of this framework aims to address the issue of insufficient instruction guidance by applying the multimodal large model MLLM (Multimodal Large Language Model) to the field of image editing.
It is reported that the MLLM model learns to obtain concise instruction expressions and provides clear visual guidance for image editing. Through end-to-end training, this model can synchronously update and utilize the latent imagination of the desired target to perform image editing. Guided by human instructions, MGIE can carry out Photoshop-style modifications, global photo optimization, and local object editing, providing users with more flexible and efficient image editing tools.
This open-source image editing framework has attracted wide attention in the industry. Experts believe that the launch of MGIE will greatly promote the development of image editing technology, offering users a more intelligent and convenient editing approach. The collaboration between Apple and UCSB once again highlights the importance of interdisciplinary cooperation, blending expertise from different fields to drive continuous technological innovation.
In the future, with the continuous optimization and popularization of the MGIE framework, it is believed that the field of image editing will see more innovation and breakthroughs, bringing users a more diverse editing experience. This news once again demonstrates the continuous progress of technology, which will continue to bring convenience and enjoyment to people’s lives.
【来源】https://www.jiqizhixin.com/articles/2024-02-05-10
Views: 4