Saved in:
Bibliographic Details
Main Authors: Zhang, Xi, Zhu, Hanwei, Zhong, Yan, Wang, Jiamang, Lin, Weisi
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.21366
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911577796509696
author Zhang, Xi
Zhu, Hanwei
Zhong, Yan
Wang, Jiamang
Lin, Weisi
author_facet Zhang, Xi
Zhu, Hanwei
Zhong, Yan
Wang, Jiamang
Lin, Weisi
contents In this work, we propose a novel framework to enable diffusion models to adapt their generation quality based on real-time network bandwidth constraints. Traditional diffusion models produce high-fidelity images by performing a fixed number of denoising steps, regardless of downstream transmission limitations. However, in practical cloud-to-device scenarios, limited bandwidth often necessitates heavy compression, leading to loss of fine textures and wasted computation. To address this, we introduce a joint end-to-end training strategy where the diffusion model is conditioned on a target quality level derived from the available bandwidth. During training, the model learns to adaptively modulate the denoising process, enabling early-stop sampling that maintains perceptual quality appropriate to the target transmission condition. Our method requires minimal architectural changes and leverages a lightweight quality embedding to guide the denoising trajectory. Experimental results demonstrate that our approach significantly improves the visual fidelity of bandwidth-adapted generations compared to naive early-stopping, offering a promising solution for efficient image delivery in bandwidth-constrained environments. Code is available at: https://github.com/xzhang9308/BADiff.
format Preprint
id arxiv_https___arxiv_org_abs_2510_21366
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle BADiff: Bandwidth Adaptive Diffusion Model
Zhang, Xi
Zhu, Hanwei
Zhong, Yan
Wang, Jiamang
Lin, Weisi
Computer Vision and Pattern Recognition
Machine Learning
In this work, we propose a novel framework to enable diffusion models to adapt their generation quality based on real-time network bandwidth constraints. Traditional diffusion models produce high-fidelity images by performing a fixed number of denoising steps, regardless of downstream transmission limitations. However, in practical cloud-to-device scenarios, limited bandwidth often necessitates heavy compression, leading to loss of fine textures and wasted computation. To address this, we introduce a joint end-to-end training strategy where the diffusion model is conditioned on a target quality level derived from the available bandwidth. During training, the model learns to adaptively modulate the denoising process, enabling early-stop sampling that maintains perceptual quality appropriate to the target transmission condition. Our method requires minimal architectural changes and leverages a lightweight quality embedding to guide the denoising trajectory. Experimental results demonstrate that our approach significantly improves the visual fidelity of bandwidth-adapted generations compared to naive early-stopping, offering a promising solution for efficient image delivery in bandwidth-constrained environments. Code is available at: https://github.com/xzhang9308/BADiff.
title BADiff: Bandwidth Adaptive Diffusion Model
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2510.21366