Saved in:
Bibliographic Details
Main Authors: Lee, Phillip Y., Sung, Minhyuk
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2403.13589
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916329940844544
author Lee, Phillip Y.
Sung, Minhyuk
author_facet Lee, Phillip Y.
Sung, Minhyuk
contents When an image generation process is guided by both a text prompt and spatial cues, such as a set of bounding boxes, do these elements work in harmony, or does one dominate the other? Our analysis of a pretrained image diffusion model that integrates gated self-attention into the U-Net reveals that spatial grounding often outweighs textual grounding due to the sequential flow from gated self-attention to cross-attention. We demonstrate that such bias can be significantly mitigated without sacrificing accuracy in either grounding by simply rewiring the network architecture, changing from sequential to parallel for gated self-attention and cross-attention. This surprisingly simple yet effective solution does not require any fine-tuning of the network but significantly reduces the trade-off between the two groundings. Our experiments demonstrate significant improvements from the original GLIGEN to the rewired version in the trade-off between textual grounding and spatial grounding.
format Preprint
id arxiv_https___arxiv_org_abs_2403_13589
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle ReGround: Improving Textual and Spatial Grounding at No Cost
Lee, Phillip Y.
Sung, Minhyuk
Computer Vision and Pattern Recognition
When an image generation process is guided by both a text prompt and spatial cues, such as a set of bounding boxes, do these elements work in harmony, or does one dominate the other? Our analysis of a pretrained image diffusion model that integrates gated self-attention into the U-Net reveals that spatial grounding often outweighs textual grounding due to the sequential flow from gated self-attention to cross-attention. We demonstrate that such bias can be significantly mitigated without sacrificing accuracy in either grounding by simply rewiring the network architecture, changing from sequential to parallel for gated self-attention and cross-attention. This surprisingly simple yet effective solution does not require any fine-tuning of the network but significantly reduces the trade-off between the two groundings. Our experiments demonstrate significant improvements from the original GLIGEN to the rewired version in the trade-off between textual grounding and spatial grounding.
title ReGround: Improving Textual and Spatial Grounding at No Cost
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2403.13589