- 博客(66)
- 收藏
- 关注
RSS订阅
原创 Computer Vision Arxiv Daily 2025.02.07
We introduce ConceptAttention, a novel method that leverages the expressive power of DiT attention layers to generate high-quality saliency maps that precisely locate textual concepts within images. Without requiring additional training, ConceptAttention r
2025-02-08 12:30:29
1055
原创 Computer Vision Arxiv Daily 2025.01.17
We introduce SynthLight, a diffusion model for portrait relighting. Our approach frames image relighting as a re-rendering problem, where pixels are transformed in response to changes in environmental lighting conditions. Using a physically-based rendering
2025-01-22 10:15:47
1255
原创 Computer Vision Arxiv Daily 2025.01.16
Video generation has achieved remarkable progress with the introduction of diffusion models, which have significantly improved the quality of generated videos. In this paper, we initially investigate the characteristics of features in intermediate layers,
2025-01-17 09:52:08
866
原创 Computer Vision Arxiv Daily 2025.01.14
Image matching, which aims to identify corresponding pixel locations between images, is crucial in a wide range of scientific disciplines, aiding in image registration, fusion, and analysis. We propose a large-scale pre-training framework that utilizes syn
2025-01-15 09:57:17
1188
2
原创 044_SSS_Counting Guidance for High Fidelity Text-to-Image Synthesis
现有的Stable Diffusion在文本生成图像时不能准确的生成指定数量的物体。
2023-07-11 10:30:01
417
原创 039_SSS_ArtFusion: Controllable Arbitrary Style Transfer using Dual Conditional LDM
本文提出了一种基于Latent Diffusion Model的un-paired风格迁移方法。
2023-06-25 09:34:18
673
原创 038_SSS_Multi-Architecture Multi-Expert Diffusion Models
本文提出了一种在diffusion的不同步数采用不同的网络结构的方法提高生成质量和效率。Diffusion模型需要大量的计算时间成本，改进方式主要有两个方面：（1）减少采样步数（2）降低网络规模现有的工作更多的关注于减少采样步。本文旨在降低网络规模。原始的Diffusion模型因为要建模不同步数，不同的噪声尺度下的特征，因此模型需要大量的参数。并且Diffusion更倾向于先学到低频的信息，然后逐渐学到高频的信息。
2023-06-11 22:07:38
767
原创 036_SS_Unsupervised Compositional Concepts Discovery with Text-to-Image Generative Models
本文提出了一种利用Diffusion模型，无监督的从少量图片中提取共同的构图概念的方法。大多数现有的concept discovery方法都集中在发现代表单个概念的潜在向量或方向，但需要监督数据标记每个concept。其他方法只关注图片中的物体对象，而不关注图片风格等信息。这项工作的贡献：(1) 本文提出了一种可扩展的方法，可以使用现有的生成模型在真实图像中无监督的发现构图概念。
2023-06-11 22:05:04
383
原创 037_SS_SyncDiffusion: Coherent Montage via Synchronized Joint Diffusions
本文提出了一种即插即用的用Diffusion生成全景图的方法。Diffusion模型通常只能生成固定大小的图像，为了生成分辨率比较高的全景图。现有的方法分成两类：第一类方法是利用Inpainting的原理，给定一部分图像补全令一部分图像。但是这种方法不能做到无缝的生成高分图像，并且很容易重复相似的图像内容。第二类方法是joint diffusion，也就是在反向采样的过程中，同时采样多个视野部分的图像，然后每一步将这些图像重叠的部分取平均值。
2023-06-11 22:04:32
623
原创 035_SS_Prompt Tuning Inversion for Text-Driven Image Editing Using Diffusion Models
本文要利用Diffusion实现高保真高质量的文本-图像编辑，也就是既保证editability，又要保证fidelity。前者要求编辑后的图像应该包含与目标提示中提供的相应文本内容良好对齐的视觉内容，而后者期望编辑部分以外的区域应尽可能接近输入图像的区域。然而，大多数方法缺乏以下之一：用户友好性（例如，需要额外的掩码或输入图像的精确描述）、对更大域的泛化或对输入图像的高保真度。
2023-06-11 22:01:09
528
原创 033_SS_Inversion-Based Creativity Transfer with Diffusion Models
033_SS_Inversion-Based Creativity Transfer with Diffusion Models
2023-02-25 09:13:08
994
原创 031_SSS_Imagic Text-Based Real Image Editing with Diffusion Models
Imagic: Text-Based Real Image Editing with Diffusion Models
2023-02-14 17:52:07
1871
原创 030_SSS_MaskSketch Unpaired Structure-guided Masked Image Generation
MaskSketch: Unpaired Structure-guided Masked Image Generation
2023-02-09 12:02:50
816
原创 029_SSS_MaskGIT Masked Generative Image Transformer（CVPR2022）
029_SSS_MaskGIT Masked Generative Image Transformer（CVPR2022）
2023-02-08 16:37:35
1640
原创 028_SSS_Fine-tuning Diffusion Models with Limited Data
Fine-tuning Diffusion Models with Limited Data
2023-02-08 16:35:07
760
原创 027_SSS_Direct Inversion Optimization-Free Text-Driven Real Image Editing with Diffusion Models
Direct Inversion: Optimization-Free Text-Driven Real Image Editing with Diffusion Models
2023-02-08 16:30:52
606
原创 026_SS_MoFusion A Framework for Denoising-Diffusion-based Motion Synthesis
MoFusion: A Framework for Denoising-Diffusion-based Motion Synthesis
2022-12-29 11:15:52
481
原创 025_SSS_BeLFusion: Latent Diffusion for Behavior-Driven Human Motion Prediction
025_SSS_BeLFusion: Latent Diffusion for Behavior-Driven Human Motion Prediction
2022-12-12 15:05:33
684
原创 024_SSS_Occupancy Flow: 4D Reconstruction by Learning Particle Dynamics(ICCV2019)
024_SSS_Occupancy Flow: 4D Reconstruction by Learning Particle Dynamics(ICCV2019)
2022-12-11 14:38:05
444
原创 023_SSS_Neural 3D Video Synthesis from Multi-view Video(CVPR2022)
023_SSS_Neural 3D Video Synthesis from Multi-view Video(CVPR2022)
2022-12-10 18:48:02
863
原创 022_SSS_Novel View Synthesis with Diffusion Models
Novel View Synthesis with Diffusion Models
2022-11-21 10:28:38
980
原创 021_SSSS_Diffusion Models Already Have a Semantic Latent Space
Diffusion Models Already Have a Semantic Latent Space
2022-11-20 14:59:07
1788
原创 007_补充_ Pytorch 反向传播和Neural ODE的反向传播
007_补充_ Pytorch 反向传播和Neural ODE的反向传播
2022-11-16 16:28:29
1006
1
原创 020_SSSS_A Style-Based Generator Architecture for Generative Adversarial Networks(StyleGAN)
A Style-Based Generator Architecture for Generative Adversarial Networks(StyleGAN)
2022-10-31 15:26:34
566
原创 019_SSSS_High-Resolution Image Synthesis and Semantic Manipulation with Conditioanl GANs
High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs(Pix2PixHD)
2022-10-30 21:09:12
932
原创 018_SS_High-Resolution Image Editing via Multi-Stage Blended Diffusion
High-Resolution Image Editing via Multi-Stage Blended Diffusion
2022-10-29 09:49:40
769
原创 017_SSS_Semantic Image Synthesis via Diffusion Models
017_SSS_Semantic Image Synthesis via Diffusion Models
2022-10-21 21:49:56
1012
原创 016 _SSS_ GAN Inversion for Consistent Video Interpolation and Manipulation
016 _SSS_ GAN Inversion for Consistent Video Interpolation and Manipulation
2022-10-18 19:27:38
493
原创 015_SSSSS_ Arbitrary Style Transfer in Real-time with Adaptive Instance Normalization
Arbitrary Style Transfer in Real-time with Adaptive Instance Normalization阅读笔记
2022-09-15 12:35:50
406
原创 014_SSS_High-Resolution Image Synthesis with Latent Diffusion Models
High-Resolution Image Synthesis with Latent Diffusion Models阅读笔记
2022-09-13 17:27:03
3095
原创 013_SSS_ Frido Feature Pyramid Diffusion for Complex Scene Image Synthesis
Frido Feature Pyramid Diffusion for Complex Scene Image Synthesis阅读笔记
2022-09-05 12:26:17
1055
原创 008_SSSS_ Improved Denoising Diffusion Probabilistic Models
Improved Denoising Diffusion Probabilistic Models 阅读笔记
2022-08-30 19:37:47
2305
原创 012_SSS_ Improving Diffusion Model Efficiency Through Patching
Improving Diffusion Model Efficiency Through Patching 阅读笔记
2022-08-04 21:17:43
603
原创 007_SSSSS_ Neural Ordinary Differential Equtions
Neural Ordinary Differential Equtions阅读笔记
2022-07-18 18:32:56
860
空空如也
空空如也
TA创建的收藏夹 TA关注的收藏夹
TA关注的人