MINDiff: Mask-Integrated Negative Attention for Controlling Overfitting in Text-to-Image Personalization

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Jeong, Seulgi, Kim, Jaeil
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911280462299136
author Jeong, Seulgi
Kim, Jaeil
author_facet Jeong, Seulgi
Kim, Jaeil
contents In the personalization process of large-scale text-to-image models, overfitting often occurs when learning specific subject from a limited number of images. Existing methods, such as DreamBooth, mitigate this issue through a class-specific prior-preservation loss, which requires increased computational cost during training and limits user control during inference time. To address these limitations, we propose Mask-Integrated Negative Attention Diffusion (MINDiff). MINDiff introduces a novel concept, negative attention, which suppresses the subject's influence in masked irrelevant regions. We achieve this by modifying the cross-attention mechanism during inference. This enables semantic control and improves text alignment by reducing subject dominance in irrelevant regions. Additionally, during the inference time, users can adjust a scale parameter lambda to balance subject fidelity and text alignment. Our qualitative and quantitative experiments on DreamBooth models demonstrate that MINDiff mitigates overfitting more effectively than class-specific prior-preservation loss. As our method operates entirely at inference time and does not alter the model architecture, it can be directly applied to existing DreamBooth models without re-training. Our code is available at https://github.com/seuleepy/MINDiff.
format Preprint
id arxiv_https___arxiv_org_abs_2511_17888
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MINDiff: Mask-Integrated Negative Attention for Controlling Overfitting in Text-to-Image Personalization
Jeong, Seulgi
Kim, Jaeil
Computer Vision and Pattern Recognition
In the personalization process of large-scale text-to-image models, overfitting often occurs when learning specific subject from a limited number of images. Existing methods, such as DreamBooth, mitigate this issue through a class-specific prior-preservation loss, which requires increased computational cost during training and limits user control during inference time. To address these limitations, we propose Mask-Integrated Negative Attention Diffusion (MINDiff). MINDiff introduces a novel concept, negative attention, which suppresses the subject's influence in masked irrelevant regions. We achieve this by modifying the cross-attention mechanism during inference. This enables semantic control and improves text alignment by reducing subject dominance in irrelevant regions. Additionally, during the inference time, users can adjust a scale parameter lambda to balance subject fidelity and text alignment. Our qualitative and quantitative experiments on DreamBooth models demonstrate that MINDiff mitigates overfitting more effectively than class-specific prior-preservation loss. As our method operates entirely at inference time and does not alter the model architecture, it can be directly applied to existing DreamBooth models without re-training. Our code is available at https://github.com/seuleepy/MINDiff.
title MINDiff: Mask-Integrated Negative Attention for Controlling Overfitting in Text-to-Image Personalization
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.17888