AttnDreamBooth: Towards Text-Aligned Personalized Text-to-Image Generation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Pang, Lianyu, Yin, Jian, Zhao, Baoquan, Wu, Feize, Wang, Fu Lee, Li, Qing, Mao, Xudong
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866929377352089600
author Pang, Lianyu
Yin, Jian
Zhao, Baoquan
Wu, Feize
Wang, Fu Lee
Li, Qing
Mao, Xudong
author_facet Pang, Lianyu
Yin, Jian
Zhao, Baoquan
Wu, Feize
Wang, Fu Lee
Li, Qing
Mao, Xudong
contents Recent advances in text-to-image models have enabled high-quality personalized image synthesis of user-provided concepts with flexible textual control. In this work, we analyze the limitations of two primary techniques in text-to-image personalization: Textual Inversion and DreamBooth. When integrating the learned concept into new prompts, Textual Inversion tends to overfit the concept, while DreamBooth often overlooks it. We attribute these issues to the incorrect learning of the embedding alignment for the concept. We introduce AttnDreamBooth, a novel approach that addresses these issues by separately learning the embedding alignment, the attention map, and the subject identity in different training stages. We also introduce a cross-attention map regularization term to enhance the learning of the attention map. Our method demonstrates significant improvements in identity preservation and text alignment compared to the baseline methods.
format Preprint
id arxiv_https___arxiv_org_abs_2406_05000
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle AttnDreamBooth: Towards Text-Aligned Personalized Text-to-Image Generation
Pang, Lianyu
Yin, Jian
Zhao, Baoquan
Wu, Feize
Wang, Fu Lee
Li, Qing
Mao, Xudong
Computer Vision and Pattern Recognition
Recent advances in text-to-image models have enabled high-quality personalized image synthesis of user-provided concepts with flexible textual control. In this work, we analyze the limitations of two primary techniques in text-to-image personalization: Textual Inversion and DreamBooth. When integrating the learned concept into new prompts, Textual Inversion tends to overfit the concept, while DreamBooth often overlooks it. We attribute these issues to the incorrect learning of the embedding alignment for the concept. We introduce AttnDreamBooth, a novel approach that addresses these issues by separately learning the embedding alignment, the attention map, and the subject identity in different training stages. We also introduce a cross-attention map regularization term to enhance the learning of the attention map. Our method demonstrates significant improvements in identity preservation and text alignment compared to the baseline methods.
title AttnDreamBooth: Towards Text-Aligned Personalized Text-to-Image Generation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2406.05000