arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2304.03373cs.CV
收藏

基于交叉注意力引导的免训练布局控制

Training-Free Layout Control with Cross-Attention Guidance

University of Oxford(牛津大学)

Minghao Chen, Iro Laina, Andrea Vedaldi

AI总结:

本文提出一种免训练的扩散模型布局控制方法,通过操控交叉注意力层实现前向与后向引导,并证明后向引导在多个基准上优于前向引导及已有工作。

摘要:

近年来基于扩散的生成器能够根据文本提示生成高质量图像。然而,它们常常忽略那些指定构图空间布局的文本指令。我们提出了一种简单的方法,无需对图像生成器进行训练或微调即可实现稳健的布局控制。我们的技术操控模型用于衔接文本与视觉信息的交叉注意力层,并在给定例如用户指定布局的情况下,将生成过程引导至期望方向。为了确定如何最好地引导注意力,我们研究了注意力图的作用,并探索了两种替代策略:前向引导和后向引导。我们在三个基准上全面评估了我们的方法,并提供了若干定性示例以及对两种策略的比较分析,证明后向引导相较于前向引导以及先前工作具有优越性。我们进一步通过将布局引导扩展到编辑真实图像的布局和上下文等应用,展示了其通用性。

英文摘要:

Recent diffusion-based generators can produce high-quality images from textual prompts. However, they often disregard textual instructions that specify the spatial layout of the composition. We propose a simple approach that achieves robust layout control without the need for training or fine-tuning of the image generator. Our technique manipulates the cross-attention layers that the model uses to interface textual and visual information and steers the generation in the desired direction given, e.g., a user-specified layout. To determine how to best guide attention, we study the role of attention maps and explore two alternative strategies, forward and backward guidance. We thoroughly evaluate our approach on three benchmarks and provide several qualitative examples and a comparative analysis of the two strategies that demonstrate the superiority of backward guidance compared to forward guidance, as well as prior work. We further demonstrate the versatility of layout guidance by extending it to applications such as editing the layout and context of real images.

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。

↑