文献类型：会议论文

英文题名：Attention Calibration for Disentangled Text-to-Image Personalization

作者：Zhang, Yanbing[1,2];Yang, Mengping[1,2];Zhou, Qin[1,2];Wang, Zhe[1,2]

机构：[1]ECUST, Dept Comp Sci & Engn, Shanghai, Peoples R China;[2]ECUST, Key Lab Smart Mfg Energy Chem Proc, Shanghai, Peoples R China

会议论文集：IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

会议日期：JUN 16-22, 2024

会议地点：Seattle, WA

语种：英文

摘要：Recent thrilling progress in large-scale text-to-image (T2I) models has unlocked unprecedented synthesis quality of AI-generated content (AIGC) including image generation, 3D and video composition. Further, personalized techniques enable appealing customized production of a novel concept given only several images as reference. However, an intriguing problem persists: Is it possible to capture multiple, novel concepts from one single reference image? In this paper, we identify that existing approaches fail to preserve visual consistency with the reference image and eliminate cross-influence from concepts. To alleviate this, we propose an attention calibration mechanism to improve the concept-level understanding of the T2I model. Specifically, we first introduce new learnable modifiers bound with classes to capture attributes of multiple concepts. Then, the classes are separated and strengthened following the activation of the cross-attention operation, ensuring comprehensive and self-contained concepts. Additionally, we suppress the attention activation of different classes to mitigate mutual influence among concepts. Together, our proposed method, dubbed DisenDiff, can learn disentangled multiple concepts from one single image and produce novel customized images with learned concepts. We demonstrate that our method outperforms the current state of the art in both qualitative and quantitative evaluations. More importantly, our proposed techniques are compatible with LoRA and in-painting pipelines, enabling more interactive experiences.