Search research articles

关于 JoVE

概览领导团队博客 JoVE 帮助中心

作者

出版流程编辑委员会范围与政策同行评审常见问题投稿

图书馆员

用户评价订阅访问资源图书馆顾问委员会常见问题

研究

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments 存档

教育

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual 教师资源中心教师网站

使用条款与条件

相关概念视频

您也可能阅读

相关文章

通过共同作者、期刊和引用图与本文相关的文章。

排序

Same author

RadGazeGen: radiomics and gaze-guided chest X-ray generation using diffusion models.

Journal of medical imaging (Bellingham, Wash.)·2026

Same author

Topology-Aware Segmentation Using Discrete Morse Theory.

... International Conference on Learning Representations·2026

Same author

TOPODIFFUSIONNET: A TOPOLOGY-AWARE DIFFUSION MODEL.

... International Conference on Learning Representations·2026

Same authorSame journal

Unifying Top-down and Bottom-up Scanpath Prediction Using Transformers.

Proceedings. IEEE Computer Society Conference on Computer Vision and Pattern Recognition·2026

Same author

Look Hear: Gaze Prediction for Speech-directed Human Attention.

Computer vision - ECCV ... : ... European Conference on Computer Vision : proceedings. European Conference on Computer Vision·2026

Same author

<i>GAZE2REPORT</i>: RADIOLOGY REPORT GENERATION VIA VISUAL-GAZE PROMPT TUNING OF LLMS.

ArXiv·2026

Same journal

CARL: A Framework for Equivariant Image Registration.

Proceedings. IEEE Computer Society Conference on Computer Vision and Pattern Recognition·2026

Same journal

Latent Drifting in Diffusion Models for Counterfactual Medical Image Synthesis.

Proceedings. IEEE Computer Society Conference on Computer Vision and Pattern Recognition·2026

Same journal

The Language of Motion: Unifying Verbal and Non-verbal Language of 3D Human Motion.

Proceedings. IEEE Computer Society Conference on Computer Vision and Pattern Recognition·2026

Same journal

Perceptual Inductive Bias Is What You Need Before Contrastive Learning.

Proceedings. IEEE Computer Society Conference on Computer Vision and Pattern Recognition·2026

Same journal

MultiMorph: On-demand Atlas Construction.

Proceedings. IEEE Computer Society Conference on Computer Vision and Pattern Recognition·2026

查看所有相关文章

Search research articles

相关实验视频

Updated: Jun 6, 2025

Image-guided Convection-enhanced Delivery into Agarose Gel Models of the Brain

Image-guided Convection-enhanced Delivery into Agarose Gel Models of the Brain

Published on: May 14, 2014

学习表现引导的扩散模型用于大图像生成.

Alexandros Graikos¹, Srikar Yellapragada¹, Minh-Quan Le¹

¹Stony Brook University.

Proceedings. IEEE Computer Society Conference on Computer Vision and Pattern Recognition

|November 28, 2024

概括

此摘要是机器生成的。

这项研究引入了一种新的方法,使用自主监督学习 (SSL) 嵌入来指导扩散模型,以生成高质量的基因病理学和卫星图像. 这种方法绕过了大量手动注释的需要,改善了图像合成和下游分类任务.

更多相关视频

Analyzing Mitochondrial Morphology Through Simulation Supervised Learning

Analyzing Mitochondrial Morphology Through Simulation Supervised Learning

Published on: March 3, 2023

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

相关实验视频

Last Updated: Jun 6, 2025

Image-guided Convection-enhanced Delivery into Agarose Gel Models of the Brain

Image-guided Convection-enhanced Delivery into Agarose Gel Models of the Brain

Published on: May 14, 2014

Analyzing Mitochondrial Morphology Through Simulation Supervised Learning

Analyzing Mitochondrial Morphology Through Simulation Supervised Learning

Published on: March 3, 2023

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

科学领域:

计算机视觉计算机视觉
机器学习机器学习
医疗成像医学成像
遥感遥感遥感遥感

背景情况:

扩散模型需要辅助数据来进行高保真样本合成,这往往是不切实际的,因为在专业领域需要大量的注释.
自主监督学习 (SSL) 表示捕获丰富的语义和视觉信息,可能作为微细的人类标签的代理.

研究的目的:

为培训基于SSL嵌入的扩散模型开发一种新的方法.
为了证明SSL嵌入的有效性作为图像生成中手动注释的替代品.
为了实现大,空间一致的图像的合成,并探索文本到图像生成.

主要方法:

训练有素的扩散模型使用自主监督学习 (SSL) 模型的嵌入.
从SSL功能生成高质量的组织病理学和遥感图像.
从SSL嵌入中组装空间一致的补丁,以构建更大的图像,保留远程依赖.
通过从文本描述生成病理学和卫星图像来证明文本对大图像的合成.

主要成果:

基于SSL嵌入条件的扩散模型成功生成了高保真基因病理学和遥感图像.
生成的图像增强了真实数据,改善了补丁级和图像规模分类的下游分类器准确性.
这些模型表现出强度和通用性,在未见的数据集上表现良好.
从文本描述中成功合成了大图像,展示了一个新的文本到大图像范式.

结论:

SSL表示是细粒度标签的有效代理,可以在没有手动注释的情况下进行扩散模型训练.
拟议的方法提高了图像合成质量,并提高了下游分类性能.
该方法是通用和强大的,在需要高保真图像生成的各种领域有潜在的应用.