苹果与加州大学圣塔芭芭拉分校（UCSB）的研究人员联合推出了一种名为MGIE（MLLM-Guided Image Editing）的图像编辑框架。该框架利用了多模态大模型MLLM来解决指令引导不足的问题。通过学习，MLLM能够生成简洁的指令表达，并提供明确的视觉相关引导。在端到端的训练过程中，扩散模型会同步更新，并利用预期目标的潜在想象力执行图像编辑。在人类指令的引导下，MGIE可以实现Photoshop风格的修改、全局照片优化以及局部对象修改。这一技术的发布对于图像编辑领域无疑是一种创新，有望推动图像编辑技术的发展。
Title: Apple and UCSB Introduce New Image Editing Technology
Keywords: Image Editing, AI Technology, Open-Source Framework
News content:
Apple and researchers from the University of California, Santa Barbara (UCSB) have jointly introduced a new image editing framework called MGIE (MLLM-Guided Image Editing). This framework uses a multi-modal large model MLLM to address the issue of insufficient guidance from instructions. Through learning, MLLM can generate concise instructions and provide clear visual guidance. During end-to-end training, the diffusion model synchronously updates and uses the potential imagination of the target to perform image editing. With human guidance, MGIE can perform Photoshop-style modifications, global photo optimization, and local object modifications. The release of this technology is undoubtedly an innovation in the field of image editing and is expected to drive the development of image editing technology.
【来源】https://www.jiqizhixin.com/articles/2024-02-05-10
Views: 2