Salvato in:
Dettagli Bibliografici
Autori principali: Liu, Shengyuan, Wang, Bo, Ma, Ye, Yang, Te, Cao, Xipeng, Chen, Quan, Li, Han, Dong, Di, Jiang, Peng
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2405.06948
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908525896138752
author Liu, Shengyuan
Wang, Bo
Ma, Ye
Yang, Te
Cao, Xipeng
Chen, Quan
Li, Han
Dong, Di
Jiang, Peng
author_facet Liu, Shengyuan
Wang, Bo
Ma, Ye
Yang, Te
Cao, Xipeng
Chen, Quan
Li, Han
Dong, Di
Jiang, Peng
contents Existing subject-driven text-to-image generation models suffer from tedious fine-tuning steps and struggle to maintain both text-image alignment and subject fidelity. For generating compositional subjects, it often encounters problems such as object missing and attribute mixing, where some subjects in the input prompt are not generated or their attributes are incorrectly combined. To address these limitations, we propose a subject-driven generation framework and introduce training-free guidance to intervene in the generative process during inference time. This approach strengthens the attention map, allowing for precise attribute binding and feature injection for each subject. Notably, our method exhibits exceptional zero-shot generation ability, especially in the challenging task of compositional generation. Furthermore, we propose a novel metric GroundingScore to evaluate subject alignment thoroughly. The obtained quantitative results serve as compelling evidence showcasing the effectiveness of our proposed method. The code will be released soon.
format Preprint
id arxiv_https___arxiv_org_abs_2405_06948
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Training-free Subject-Enhanced Attention Guidance for Compositional Text-to-image Generation
Liu, Shengyuan
Wang, Bo
Ma, Ye
Yang, Te
Cao, Xipeng
Chen, Quan
Li, Han
Dong, Di
Jiang, Peng
Computer Vision and Pattern Recognition
Existing subject-driven text-to-image generation models suffer from tedious fine-tuning steps and struggle to maintain both text-image alignment and subject fidelity. For generating compositional subjects, it often encounters problems such as object missing and attribute mixing, where some subjects in the input prompt are not generated or their attributes are incorrectly combined. To address these limitations, we propose a subject-driven generation framework and introduce training-free guidance to intervene in the generative process during inference time. This approach strengthens the attention map, allowing for precise attribute binding and feature injection for each subject. Notably, our method exhibits exceptional zero-shot generation ability, especially in the challenging task of compositional generation. Furthermore, we propose a novel metric GroundingScore to evaluate subject alignment thoroughly. The obtained quantitative results serve as compelling evidence showcasing the effectiveness of our proposed method. The code will be released soon.
title Training-free Subject-Enhanced Attention Guidance for Compositional Text-to-image Generation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2405.06948