ByteEdit：提升、遵循并加速生成式图像编辑
ByteEdit: Boost, Comply and Accelerate Generative Image Editing
AI总结：
本文提出 ByteEdit 反馈学习框架，融合图像与像素级奖励模型及对抗式渐进反馈策略，显著提升图像扩图/补绘的质量、一致性、指令遵循与推理效率。
摘要：
近期基于扩散模型的生成式图像编辑进展引发了深刻变革，重塑了图像扩图和补绘任务的格局。尽管取得了这些进展，该领域仍面临固有挑战，包括：i）质量较差；ii）一致性不佳；iii）指令遵循不足；iv）生成效率未达最优。为应对这些障碍，我们提出 ByteEdit，这是一个精心设计的创新反馈学习框架，旨在提升、遵循并加速生成式图像编辑任务。ByteEdit 无缝集成了专注于增强美学和图文对齐的图像奖励模型，同时引入了一个稠密的像素级奖励模型，专门用于促进输出的连贯性。此外，我们提出了一种开创性的对抗式渐进反馈学习策略，以加快模型的推理速度。通过广泛的大规模用户评估，我们证明 ByteEdit 在生成质量和一致性方面均超越了包括 Adobe、Canva 和美图在内的主流生成式图像编辑产品。与基线模型相比，ByteEdit-Outpainting 在质量和一致性方面分别展现出 388% 和 135% 的显著提升。实验还验证了我们的加速模型在质量和一致性方面保持了出色的性能结果。
英文摘要：
Recent advancements in diffusion-based generative image editing have sparked a profound revolution, reshaping the landscape of image outpainting and inpainting tasks. Despite these strides, the field grapples with inherent challenges, including: i) inferior quality; ii) poor consistency; iii) insufficient instrcution adherence; iv) suboptimal generation efficiency. To address these obstacles, we present ByteEdit, an innovative feedback learning framework meticulously designed to Boost, Comply, and Accelerate Generative Image Editing tasks. ByteEdit seamlessly integrates image reward models dedicated to enhancing aesthetics and image-text alignment, while also introducing a dense, pixel-level reward model tailored to foster coherence in the output. Furthermore, we propose a pioneering adversarial and progressive feedback learning strategy to expedite the model's inference speed. Through extensive large-scale user evaluations, we demonstrate that ByteEdit surpasses leading generative image editing products, including Adobe, Canva, and MeiTu, in both generation quality and consistency. ByteEdit-Outpainting exhibits a remarkable enhancement of 388% and 135% in quality and consistency, respectively, when compared to the baseline model. Experiments also verfied that our acceleration models maintains excellent performance results in terms of quality and consistency.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
