Saved in:
Bibliographic Details
Main Authors: Li, Qinchan, Chen, Kenneth, Su, Changyue, Sun, Qi
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2412.05780
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910759853752320
author Li, Qinchan
Chen, Kenneth
Su, Changyue
Sun, Qi
author_facet Li, Qinchan
Chen, Kenneth
Su, Changyue
Sun, Qi
contents Diffusion models have shown unprecedented success in the task of text-to-image generation. While these models are capable of generating high-quality and realistic images, the complexity of sequential denoising has raised societal concerns regarding high computational demands and energy consumption. In response, various efforts have been made to improve inference efficiency. However, most of the existing efforts have taken a fixed approach with neural network simplification or text prompt optimization. Are the quality improvements from all denoising computations equally perceivable to humans? We observed that images from different text prompts may require different computational efforts given the desired content. The observation motivates us to present BudgetFusion, a novel model that suggests the most perceptually efficient number of diffusion steps before a diffusion model starts to generate an image. This is achieved by predicting multi-level perceptual metrics relative to diffusion steps. With the popular Stable Diffusion as an example, we conduct both numerical analyses and user studies. Our experiments show that BudgetFusion saves up to five seconds per prompt without compromising perceptual similarity. We hope this work can initiate efforts toward answering a core question: how much do humans perceptually gain from images created by a generative model, per watt of energy?
format Preprint
id arxiv_https___arxiv_org_abs_2412_05780
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle BudgetFusion: Perceptually-Guided Adaptive Diffusion Models
Li, Qinchan
Chen, Kenneth
Su, Changyue
Sun, Qi
Computer Vision and Pattern Recognition
Artificial Intelligence
Diffusion models have shown unprecedented success in the task of text-to-image generation. While these models are capable of generating high-quality and realistic images, the complexity of sequential denoising has raised societal concerns regarding high computational demands and energy consumption. In response, various efforts have been made to improve inference efficiency. However, most of the existing efforts have taken a fixed approach with neural network simplification or text prompt optimization. Are the quality improvements from all denoising computations equally perceivable to humans? We observed that images from different text prompts may require different computational efforts given the desired content. The observation motivates us to present BudgetFusion, a novel model that suggests the most perceptually efficient number of diffusion steps before a diffusion model starts to generate an image. This is achieved by predicting multi-level perceptual metrics relative to diffusion steps. With the popular Stable Diffusion as an example, we conduct both numerical analyses and user studies. Our experiments show that BudgetFusion saves up to five seconds per prompt without compromising perceptual similarity. We hope this work can initiate efforts toward answering a core question: how much do humans perceptually gain from images created by a generative model, per watt of energy?
title BudgetFusion: Perceptually-Guided Adaptive Diffusion Models
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2412.05780