图像生成 - arXivDaily 专题

1806.09502 2018-06-26 math.OC 78%

Optimal control of diffusion processes pertaining to an opioid epidemic dynamical model with random perturbations

Getachew K. Befekadu, Quanyan Zhu

专题命中可控生成：diffusion(title,abstract)

Comments 13 pages - Version 1.0 - June 25, 2018 (Additional Note: This work is, in some sense, a continuation of our previous paper arXiv:1805.12534)

1806.05953 2018-06-18 stat.ML cs.LG 78%

Controllable Semantic Image Inpainting

Jin Xu, Yee Whye Teh

专题命中可控生成：inpainting(title,abstract)

1708.07751 2017-08-28 math.PR 78%

Optimal Control of Forward-Backward Stochastic Differential System of Jump Diffusion with Observation Noise: Stochastic Maximum Principle

Qingxin Meng

专题命中可控生成：diffusion(title,abstract)

Comments arXiv admin note: substantial text overlap with arXiv:1708.03008, arXiv:1708.03004, arXiv:1708.05663

1702.00194 2017-02-02 math.OC math.PR 78%

Existence of an Optimal Control for a coupled FBSDE with a non degenerate diffusion coefficient

Khaled Bahlali, Omar Kebiri, Brahim Mezerdi, Ahmed Mtiraoui

专题命中可控生成：diffusion(title,abstract)

Comments 18 pages, Submitted for publication in Stochastics

1412.4917 2016-10-12 math.PR 78%

Tube estimates for diffusion processes under a weak Hörmander condition

Paolo Pigato

专题命中可控生成：diffusion(title,abstract)

1206.5209 2016-04-05 math.OC 78%

On the cost of null-control of an artificial advection-diffusion problem

Pierre Cornilleau, Sergio Guerrero

专题命中可控生成：diffusion(title,abstract)

Comments 16 pages

Journal ref ESAIM: Control, Optimisation and Calculus of Variations (2013) Volume: 19, Issue: 4, pages 1209-1224

1210.3739 2014-10-16 stat.ME 78%

Control Theory and Experimental Design in Diffusion Processes

Giles Hooker, Kevin K. Lin, Bruce Rogers

专题命中可控生成：diffusion(title,abstract)

1409.6731 2014-09-30 math.OC math.DS 78%

On the risk-sensitive escape control for diffusion processes pertaining to an expanding construction of distributed control systems

Getachew K. Befekadu, Panos J. Antsaklis

专题命中可控生成：diffusion(title,abstract)

Comments 11 Pages (Additional Note: This work is, in some sense, a continuation of our previous paper arXiv:1409.2751 [math.DS])

1409.2751 2014-09-12 math.DS 78%

On the minimum exit rate for a diffusion process pertaining to a chain of distributed control systems with random perturbations

Getachew K. Befekadu, Panos J. Antsaklis

专题命中可控生成：diffusion(title,abstract)

Comments 12 Pages. (Additional Note: This work is, in some sense, a continuation of our previous paper arXiv:1408.6260.)

1302.5034 2013-07-11 math.OC 78%

A Correction to "Dynkin Games Via Dirichlet Forms and Singular Control of One-Dimensional Diffusion"

Yipeng Yang

专题命中可控生成：diffusion(title,abstract)

Comments This paper has been withdrawn. The content of this paper has been added to the appendix of the author's another paper 'A Multi-dimensional Stochastic Singular Control Problem Via Dynkin Game and Dirichlet Form'

1202.4771 2012-02-23 math.PR 78%

Tubes estimates for diffusion processes under a local Hörmander condition of order one

Vlad Bally, Lucia Caramellino

专题命中可控生成：diffusion(title,abstract)

math/0602365 2009-12-01 math.PR 78%

Optimally coupling the Kolmogorov diffusion, and related optimal control problems

Kalvis M. Jansons, Paul D. Metcalfe

专题命中可控生成：diffusion(title,abstract)

Comments 19 pages, 6 figures [corrected missing references]

2606.15819 2026-06-16 cs.CV cs.AI 新提交 77%

SACE: Concept Erasure at the Semantic Singularity in Visual Autoregressive Models

SACE: 视觉自回归模型中的语义奇点概念擦除

Siya Yang, Nanxiang Jiang, Zhaoxin Fan, Yunfeng Diao

机构 * School of Computer Science and Engineering, Sun Yat-sen University（中山大学计算机科学与工程学院）； School of Computer Science and Technology, Beijing Institute of Technology（北京理工大学计算机科学与技术学院）

专题命中可控生成：text-to-image(abstract)；diffusion(abstract)；image synthesis(abstract)；分类 cs.CV

AI总结针对视觉自回归模型应用现有擦除技术导致语义崩溃和视觉伪影的问题，提出语义奇点公理并通过增量语义显著性分析验证，进而引入首个尺度感知的概念擦除框架SACE，在首尺度耦合熵正则化擦除目标与恢复性保存损失，实现精确概念擦除。

2606.14025 2026-06-15 cs.CV 新提交 77%

GarmentSketch: Large-scale Sketch-to-Fashion Benchmark

GarmentSketch：大规模草图到时尚基准

Duong-Duy-Khang Bui, Minh-Tan Pham, Tam V. Nguyen, Minh-Triet Tran, Trung-Nghia Le

专题命中可控生成：image generation(abstract)；text-to-image(abstract)；image synthesis(abstract)；分类 cs.CV

AI总结为解决时尚草图到图像合成缺乏大规模配对数据的问题，构建了包含26249对草图-文本描述的GarmentSketch数据集，并基于多模态大模型与人工精炼生成描述，评估了现有生成模型的性能。

Comments ICCCI 2026. Project page: https://khangbdd.github.io/garmentsketch

2606.06601 2026-06-08 cs.CV cs.AI cs.LG 新提交 77%

Direct 3D-Aware Object Insertion via Decomposed Visual Proxies

通过分解视觉代理实现直接3D感知物体插入

Jingbo Gong, Yikai Wang, Yushi Lan, Yuhao Wan, Ziheng Ouyang, Rui Zhao, Ming-Ming Cheng, Qibin Hou, Chen Change Loy

机构 * Google（谷歌）； Black Forest Labs（黑森林实验室）

专题命中可控生成：diffusion(abstract)；inpainting(abstract)；image synthesis(abstract)；分类 cs.CV

AI总结提出DIRECT框架，通过分解外观、几何和上下文引导，实现可控制3D姿态的物体插入，在几何可控性和视觉质量上优于现有方法。

Comments ICML 2026; Project Page: https://gong1130.github.io/DIRECT/

2604.26883 2026-04-30 cs.CV 77%

SEAL: Semantic-aware Single-image Sticker Personalization with a Large-scale Sticker-tag Dataset

SEAL: 基于大规模贴纸-标签数据集的语义感知单图贴纸个性化

Changhyun Roh, Yonghyun Jeong, Jonghyun Lee, Chanho Eom, Jihyong Oh

专题命中可控生成：image generation(abstract)；text-to-image(abstract)；diffusion(abstract)；分类 cs.CV

AI总结 SEAL通过引入语义感知的单图贴纸个性化方法，解决扩散模型在贴纸个性化中的视觉纠缠和结构刚性问题，利用StickerBench数据集实现属性级控制，提升身份保持与上下文可控性。

Comments The last two authors are co-corresponding authors. Please visit our project page at https://cmlab-korea.github.io/SEAL

2604.22847 2026-04-28 cs.CV 77%

Dream-Cubed: Controllable Generative Modeling in Minecraft by Training on Billions of Cubes

Dream-Cubed：通过训练数十亿个立方体实现Minecraft中的可控生成建模

Tim Merino, Sam Earle, Ryunosuke Iwai, Julian Togelius, Edoardo Cetin

机构 * New York University（纽约大学）； Sakana AI

专题命中可控生成：inpainting(abstract,abstract_cn)；diffusion(abstract)；分类 cs.CV

AI总结本文提出Dream-Cubed数据集及基于立方体的生成模型，用于高效生成交互式3D环境，通过分析扩散模型的不同形式和架构，评估生成质量并释放预训练模型以推动未来研究。

2604.19720 2026-04-22 cs.CV 77%

ReImagine: Rethinking Controllable High-Quality Human Video Generation via Image-First Synthesis

ReImagine：重新思考通过图像优先合成实现可控高质量人类视频生成

Zhengwentai Sun, Keru Zheng, Chenghong Li, Hongjie Liao, Xihe Yang, Heyuan Li, Yihao Zhi, Shuliang Ning, Shuguang Cui, Xiaoguang Han

机构 * School of Science and Engineering, The Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳）科学与工程学院）； Future Network of Intelligence Institute, CUHK-Shenzhen, China（CUHK-深圳智能网络研究院）

专题命中可控生成：image generation(abstract)；diffusion(abstract)；image synthesis(abstract)；分类 cs.CV

AI总结本文提出通过图像优先合成方法，结合预训练图像模型与SMPL-X运动引导，生成高质量且时间一致的视频，同时释放人类数据集与辅助模型。

2604.09405 2026-04-13 cs.CV 77%

EGLOCE: Training-Free Energy-Guided Latent Optimization for Concept Erasure

EGLOCE：无需训练的能量引导潜在优化用于概念擦除

Junyeong Ahn, Seojin Yoon, Sungyong Baik

机构 * KAIST AI（韩国科学技术院人工智能）； Hanyang University（汉阳大学）

专题命中可控生成：image generation(abstract)；text-to-image(abstract)；diffusion(abstract)；分类 cs.CV

AI总结 EGLOCE通过在推理过程中重新定向噪声潜在变量，实现无需训练的概念擦除，结合排斥能量和保留能量，提升图像质量和提示对齐性。

2603.27199 2026-03-31 cs.CV 77%

Let Triggers Control: Frequency-Aware Dropout for Effective Token Control

让触发器控制：面向频率的Dropout用于有效的标记控制

Junyoung Koh, Hoyeon Moon, Dongha Kim, Seungmin Lee, Sanghyun Park, Min Song

机构 * Yonsei University（延世大学）； Onoma AI

专题命中可控生成：image generation(abstract)；text-to-image(abstract)；diffusion(abstract)；分类 cs.CV

AI总结本文提出频率感知Dropout方法，通过分析触发器与上下文的共现关系，提升文本到图像生成中触发器的可控性与个性化能力，无需增加参数或架构修改。

Comments CVPR 2026 P13N: Personalization in Generative AI workshop

2603.21136 2026-03-24 cs.CV 77%

MS-CustomNet: Controllable Multi-Subject Customization with Hierarchical Relational Semantics

MS-CustomNet：基于层次关系语义的可控多主体定制

Pengxiang Cai, Mengyang Li

机构 * School of Information Science and Engineering, East China University of Science and Technology（信息科学与工程学院，东华大学）； Center for Applied Mathematics, Tianjin University（应用数学中心，天津大学）

专题命中可控生成：image generation(abstract)；text-to-image(abstract)；diffusion(abstract)；分类 cs.CV

AI总结本文提出MS-CustomNet，通过层次关系语义实现多主体定制，支持零样本整合多个用户提供的对象，并允许用户显式定义主体间的层级安排与空间位置，提升多主体图像生成的精细控制能力。

2511.14063 2026-03-17 cs.CV 77%

Semantic Context Matters: Improving Conditioning for Autoregressive Models

语义上下文至关重要：改进自回归模型的条件化

Dongyang Jin, Ryan Xu, Jianhao Zeng, Rui Lan, Yancheng Bai, Lei Sun, Xiangxiang Chu

专题命中可控生成：image generation(abstract)；diffusion(abstract)；image editing(abstract)；分类 cs.CV

AI总结本文提出SCAR方法，通过压缩语义前缀和语义对齐指导提升自回归模型的条件化能力，实现更高质量的图像编辑和可控生成。

2603.09759 2026-03-11 cs.CV 77%

LogoDiffuser: Training-Free Multilingual Logo Generation and Stylization via Letter-Aware Attention Control

LogoDiffuser: 无需训练的多语言标志生成与风格化通过字母感知注意力控制

Mingyu Kang, Hyein Seo, Yuna Jeong, Junhyeong Park, Yong Suk Choi

机构 * Department of Artificial Intelligence, Hanyang University（翰阳大学人工智能系）； Department of Computer Science, Hanyang University（翰阳大学计算机科学系）

专题命中可控生成：image generation(abstract)；text-to-image(abstract)；diffusion(abstract)；分类 cs.CV

AI总结 LogoDiffuser通过字母感知注意力控制，无需训练实现多语言标志生成与风格化，通过多模态扩散变换器整合字符结构与视觉设计，提升多语言标志生成效果。

2505.18612 2026-02-23 cs.CV 77%

Mod-Adapter: Tuning-Free and Versatile Multi-concept Personalization via Modulation Adapter

Mod-Adapter：通过调制适配器实现无需微调的多概念个性化

Weizhi Zhong, Huan Yang, Zheng Liu, Huiguo He, Zijian He, Xuesong Niu, Di Zhang, Guanbin Li

机构 * School of Computer Science and Engineering, Sun Yat-sen University（中山大学计算机科学与工程学院）； Kolors Team, Kuaishou Technology（快手科技Kolors团队）； Shenzhen Loop Area Institute, Shenzhen, China（深圳循环区研究所）； Guangdong Key Laboratory of Big Data Analysis and Processing, Guangzhou, China（广东大数据分析与处理重点实验室）； South China University of Technology, Guangzhou, China（华南理工大学）

专题命中可控生成：image generation(abstract)；text-to-image(abstract)；diffusion(abstract)；分类 cs.CV

AI总结本文提出Mod-Adapter方法，通过调制适配器实现无需微调的多概念个性化，有效定制物体和抽象概念，提升生成质量。

Comments Accepted by ICLR 2026, project page: https://weizhi-zhong.github.io/Mod-Adapter

2602.14443 2026-02-17 cs.CV 77%

Controlling Your Image via Simplified Vector Graphics

通过简化矢量图形控制图像

Lanqing Guo, Xi Liu, Yufei Wang, Zhihao Li, Siyu Huang

机构 * The University of Texas at Austin, USA（德克萨斯大学奥斯汀分校）； Clemson University, USA（克莱姆森大学）； SparcAI Inc., USA（SparcAI公司）

专题命中可控生成：image generation(abstract)；image editing(abstract)；image synthesis(abstract)；分类 cs.CV

AI总结本研究提出通过简化矢量图形实现图像元素层面的可控生成，通过分层结构和语义对齐的矢量图形表示，实现对几何、颜色和物体语义的精确控制，适用于图像编辑和精细内容创作。

Comments Preprint

2506.11924 2026-02-09 cs.CV 77%

Aligned Novel View Image and Geometry Synthesis via Cross-modal Attention Instillation

通过跨模态注意力安装实现对齐的新视角图像和几何合成

Min-Seop Kwak, Junho Kim, Sangdoo Yun, Dongyoon Han, Taekyung Kim, Seungryong Kim, Jin-Hwa Kim

机构 * NAVER AI Lab（NAVER AI实验室）； KAIST AI（韩国科学技术院人工智能学院）； SNU AIIS（首尔国立大学人工智能研究所）

专题命中可控生成：diffusion(abstract)；inpainting(abstract)；image synthesis(abstract)；分类 cs.CV

AI总结本文提出了一种基于扩散的框架，通过跨模态注意力蒸馏实现新视角图像和几何的对齐生成，提升几何鲁棒性和重建质量。

Comments Project page at https://cvlab-kaist.github.io/MoAI

2512.22272 2025-12-30 cs.CV 77%

Human-Aligned Generative Perception: Bridging Psychophysics and Generative Models

人类对齐的生成感知：连接心理学物理与生成模型

Antara Titikhsha, Om Kulkarni, Dharun Muthaiah

专题命中可控生成：text-to-image(abstract)；diffusion(abstract)；image synthesis(abstract)；分类 cs.CV

AI总结本文提出通过轻量级判别器引导生成模型，实现几何与风格的可控分离，提升文本到图像合成的语义对齐度。

2510.14882 2025-10-17 cs.CV 77%

ScaleWeaver: Weaving Efficient Controllable T2I Generation with Multi-Scale Reference Attention

Keli Liu, Zhendong Wang, Wengang Zhou, Shaodong Xu, Ruixiao Dong, Houqiang Li

机构 * University of Science and Technology of China（中国科学技术大学）

专题命中可控生成：image generation(abstract)；text-to-image(abstract)；diffusion(abstract)；分类 cs.CV

2509.17757 2025-09-23 cs.CV cs.MA 77%

Multi-Agent Amodal Completion: Direct Synthesis with Fine-Grained Semantic Guidance

Hongxing Fan, Lipeng Wang, Haohua Chen, Zehuan Huang, Jiangtao Wu, Lu Sheng

机构 * School of Computer Science and Engineering（计算机科学与工程学院）； Beihang University（北京航空航天大学）； School of Software（软件学院）

专题命中可控生成：diffusion(abstract)；image editing(abstract)；inpainting(abstract)；分类 cs.CV

2506.16806 2025-09-23 cs.CV 77%

FOCUS: Unified Vision-Language Modeling for Interactive Editing Driven by Referential Segmentation

Fan Yang, Yousong Zhu, Xin Li, Yufei Zhan, Hongyin Zhao, Shurong Zheng, Yaowei Wang, Ming Tang, Jinqiao Wang

机构 * Foundation Model Research Center, Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所基础模型研究中心）； School of Artificial Intelligence, University of Chinese Academy of Science（中国科学院大学人工智能学院）； Peng Cheng Laboratory, Shenzhen, China（鹏城实验室）； Wuhan AI Research, Wuhan, China（武汉人工智能研究所）

专题命中可控生成：image generation(abstract)；diffusion(abstract)；image editing(abstract)；分类 cs.CV

1. 可控生成 4201 篇

Optimal control of diffusion processes pertaining to an opioid epidemic dynamical model with random perturbations

Controllable Semantic Image Inpainting

Optimal Control of Forward-Backward Stochastic Differential System of Jump Diffusion with Observation Noise: Stochastic Maximum Principle

Existence of an Optimal Control for a coupled FBSDE with a non degenerate diffusion coefficient

Tube estimates for diffusion processes under a weak Hörmander condition

On the cost of null-control of an artificial advection-diffusion problem

Control Theory and Experimental Design in Diffusion Processes

On the risk-sensitive escape control for diffusion processes pertaining to an expanding construction of distributed control systems

On the minimum exit rate for a diffusion process pertaining to a chain of distributed control systems with random perturbations

A Correction to "Dynkin Games Via Dirichlet Forms and Singular Control of One-Dimensional Diffusion"

Tubes estimates for diffusion processes under a local Hörmander condition of order one

Optimally coupling the Kolmogorov diffusion, and related optimal control problems

SACE: Concept Erasure at the Semantic Singularity in Visual Autoregressive Models

GarmentSketch: Large-scale Sketch-to-Fashion Benchmark

Direct 3D-Aware Object Insertion via Decomposed Visual Proxies

SEAL: Semantic-aware Single-image Sticker Personalization with a Large-scale Sticker-tag Dataset

Dream-Cubed: Controllable Generative Modeling in Minecraft by Training on Billions of Cubes

ReImagine: Rethinking Controllable High-Quality Human Video Generation via Image-First Synthesis

EGLOCE: Training-Free Energy-Guided Latent Optimization for Concept Erasure

Let Triggers Control: Frequency-Aware Dropout for Effective Token Control

MS-CustomNet: Controllable Multi-Subject Customization with Hierarchical Relational Semantics

Semantic Context Matters: Improving Conditioning for Autoregressive Models

LogoDiffuser: Training-Free Multilingual Logo Generation and Stylization via Letter-Aware Attention Control

Mod-Adapter: Tuning-Free and Versatile Multi-concept Personalization via Modulation Adapter

Controlling Your Image via Simplified Vector Graphics

Aligned Novel View Image and Geometry Synthesis via Cross-modal Attention Instillation

Human-Aligned Generative Perception: Bridging Psychophysics and Generative Models

ScaleWeaver: Weaving Efficient Controllable T2I Generation with Multi-Scale Reference Attention

Multi-Agent Amodal Completion: Direct Synthesis with Fine-Grained Semantic Guidance

FOCUS: Unified Vision-Language Modeling for Interactive Editing Driven by Referential Segmentation