LoRA-Composer: Leveraging Low-Rank Adaptation for Multi-Concept Customization in Training-Free Diffusion Models
IEEE Transactions on Image Processing · 2025 · 34 · 8145-8158
DOI: 10.1109/tip.2025.3633153 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Image Processing, 2025. Verfasst von Yang Yang und 11 weiteren Autor:innen. Die Arbeit wurde bisher 31-mal zitiert (Crossref) und nennt 52 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Key Research and Development Program of Ningbo, Yongjiang Talent Introduction Program.
Abstract (Originalsprache)
Customization generation techniques have significantly advanced the synthesis of specific concepts across varied contexts. Multi-concept customization emerges as the challenging task within this domain. Existing approaches often rely on training a fusion matrix of multiple Low-Rank Adaptations (LoRAs) to merge various concepts into a single image. However, we identify this straightforward method faces two major challenges: 1) concept confusion, where the model struggles to preserve distinct individual characteristics, and 2) concept vanishing, where the model fails to generate the intended subjects. To address these issues, we introduce LoRA-Composer, a training-free framework designed for seamlessly integrating multiple LoRAs, thereby enhancing the harmony among different concepts within generated images. LoRA-Composer addresses concept vanishing through concept injection constraints, enhancing visibility via an expanded cross-attention mechanism. To combat concept confusion, concept isolation constraints are introduced, refining the self-attention computation. Furthermore, we propose two inference techniques to accelerate inference speed without performance degradation and enhance the accuracy of the generated region, respectively. Extensive experiments demonstrate that LoRA-Composer significantly outperforms standard baselines, especially in scenarios without image-based conditions such as canny edge or pose estimation.
Förderung
- National Natural Science Foundation of China
- Key Research and Development Program of Ningbo
- Yongjiang Talent Introduction Program
- China Computer Federation (CCF)-Tencent Open Fund
Zitieren
APA
Yang, Y., Wang, W., Peng, L., Song, C., Chen, Y., Li, H., Yang, X., Lu, Q., Cai, D., He, X., Wu, B., & Liu, W. (2025). LoRA-Composer: Leveraging Low-Rank Adaptation for Multi-Concept Customization in Training-Free Diffusion Models. IEEE Transactions on Image Processing, 34, 8145-8158. https://doi.org/10.1109/tip.2025.3633153
Vancouver
Yang Y, Wang W, Peng L, Song C, Chen Y, Li H, et al. LoRA-Composer: Leveraging Low-Rank Adaptation for Multi-Concept Customization in Training-Free Diffusion Models. IEEE Transactions on Image Processing. 2025;34:8145-8158. doi:10.1109/tip.2025.3633153
BibTeX
@article{Yang2025,
author = {Yang, Yang and Wang, Wen and Peng, Liang and Song, Chaotian and Chen, Yao and Li, Hengjia and Yang, Xiaolong and Lu, Qinglin and Cai, Deng and He, Xiaofei and Wu, Boxi and Liu, Wei},
title = {LoRA-Composer: Leveraging Low-Rank Adaptation for Multi-Concept Customization in Training-Free Diffusion Models},
journal = {IEEE Transactions on Image Processing},
year = {2025},
volume = {34},
pages = {8145--8158},
doi = {10.1109/tip.2025.3633153}
}
EUVIMED