En3D vs Lovart：怎么选？

关于 En3D

Papersarxiv:2401.01173Copy markdownEn3D: An Enhanced Generative Model for Sculpting 3D Humans from 2D Synthetic DataPublished on Jan 2, 2024·Submitted byAKon Jan 3, 2024Upvote12+4Authors:Yifang Men,Biwen Lei,Yuan Yao,Miaomiao Cui,Zhouhui Lian,Xuansong XieAbstractEn3D generates high-quality 3D human avatars using synthetic 2D data without relying on pre-existing 3D models, employing a combination of a 3D generator, geometry sculptor, and texturing module.Generated byQwen/Qwen2.5-Coder-32B-InstructWe present En3D, an enhancedgenerative schemefor sculpting high-quality 3D human avatars. Unlike previous works that rely on scarce 3D datasets or limited 2D collections with imbalanced viewing angles and imprecise pose priors, our approach aims to develop a zero-shot 3Dgenerative schemecapable of producing visually realistic, geometrically accurate and content-wise diverse 3D humans without relying on pre-existing 3D or 2D assets. To address this challenge, we introduce a meticulously crafted workflow that implements accurate physical modeling to learn the enhanced3D generative modelfrom synthetic 2D data. During inference, we integrate optimization modules to bridge the gap between realistic appearances and coarse 3D shapes. Specifically, En3D comprises three modules: a3D generatorthat accurately models generalizable 3D humans with realistic appearance from synthesized balanced, diverse, and structured human images; ageometry sculptorthat enhances shape quality using multi-view normal constraints for intricate human anatomy; and atexturing modulethat disentangles explicit texture maps with fidelity and editability, leveragingsemantical UV partitioningand adifferentiable rasterizer. Experimental results show that our approach significantly outperforms prior works in terms of image quality, geometry accuracy and content diversity. We also showcase the applicability of our generated avatars for animation and editing, as well as the scalability of our approach for content-style free adaptation.View arXiv pageView PDFAdd to collectionCommunitymiaoyinJan 4, 2024This comment has been hiddenmiaoyinJan 4, 2024This comment has been hiddenAmirsefatJan 19, 2024A man with a womanReplyChromaFlowFeb 10, 2024This comment has been hiddenLucas3467May 18, 2024ReplywwwguruJun 23, 2024This comment has been hiddenwwwguruJun 23, 2024Перерисуй фото в 3DReplyAnDongEluosiOct 11, 2024ReplyAditya98Jan 22, 2025convert this watch into a 3d model whichi is seen by realistic ,See translationReplyEditPreviewUpload images, audio, and videos by dragging in the text input, pasting, orclicking here.Tap or paste here to upload imagesComment·Sign uporlog into comment

AbstractEn3D generates high-quality 3D human avatars using synthetic 2D data without relying on pre-existing 3D models, employing a combination of a 3D generator, geometry sculptor, and texturing module.Generated byQwen/Qwen2.5-Coder-32B-InstructWe present En3D, an enhancedgenerative schemefor sculpting high-quality 3D human avatars. Unlike previous works that rely on scarce 3D datasets or limited 2D collections with imbalanced viewing angles and imprecise pose priors, our approach aims to develop a zero-shot 3Dgenerative schemecapable of producing visually realistic, geometrically accurate and content-wise diverse 3D humans without relying on pre-existing 3D or 2D assets. To address this challenge, we introduce a meticulously crafted workflow that implements accurate physical modeling to learn the enhanced3D generative modelfrom synthetic 2D data. During inference, we integrate optimization modules to bridge the gap between realistic appearances and coarse 3D shapes. Specifically, En3D comprises three modules: a3D generatorthat accurately models generalizable 3D humans with realistic appearance from synthesized balanced, diverse, and structured human images; ageometry sculptorthat enhances shape quality using multi-view normal constraints for intricate human anatomy; and atexturing modulethat disentangles explicit texture maps with fidelity and editability, leveragingsemantical UV partitioningand adifferentiable rasterizer. Experimental results show that our approach significantly outperforms prior works in terms of image quality, geometry accuracy and content diversity. We also showcase the applicability of our generated avatars for animation and editing, as well as the scalability of our approach for content-style free adaptation.

En3D generates high-quality 3D human avatars using synthetic 2D data without relying on pre-existing 3D models, employing a combination of a 3D generator, geometry sculptor, and texturing module.Generated byQwen/Qwen2.5-Coder-32B-InstructWe present En3D, an enhancedgenerative schemefor sculpting high-quality 3D human avatars. Unlike previous works that rely on scarce 3D datasets or limited 2D collections with imbalanced viewing angles and imprecise pose priors, our approach aims to develop a zero-shot 3Dgenerative schemecapable of producing visually realistic, geometrically accurate and content-wise diverse 3D humans without relying on pre-existing 3D or 2D assets. To address this challenge, we introduce a meticulously crafted workflow that implements accurate physical modeling to learn the enhanced3D generative modelfrom synthetic 2D data. During inference, we integrate optimization modules to bridge the gap between realistic appearances and coarse 3D shapes. Specifically, En3D comprises three modules: a3D generatorthat accurately models generalizable 3D humans with realistic appearance from synthesized balanced, diverse, and structured human images; ageometry sculptorthat enhances shape quality using multi-view normal constraints for intricate human anatomy; and atexturing modulethat disentangles explicit texture maps with fidelity and editability, leveragingsemantical UV partitioningand adifferentiable rasterizer. Experimental results show that our approach significantly outperforms prior works in terms of image quality, geometry accuracy and content diversity. We also showcase the applicability of our generated avatars for animation and editing, as well as the scalability of our approach for content-style free adaptation.

核心功能

• Join the discussion on this paper page（自动优化版）
• En3D generates high-quality 3D human avatars using synthetic 2D data without relying on pre-existing 3D models, employing a combination of a 3D generator, geometry sculptor, and texturing module.（自动优化版）
• A man with a woman（自动优化版）
• Перерисуй фото в 3D（自动优化版）
• convert this watch into a 3d model whichi is seen by realistic ,（自动优化版）
• ·Sign uporlog into comment（自动优化版）
• Get this paper in your agent:（自动优化版）
• No dataset linking this paper（自动优化版）

关于 Lovart

Design visual assets for a running shoe product page, including a hero image, product close-ups, and athlete-on-foot shots that highlight performance and materials.

Text shouldn't be baked in. Lovart separates typography into editable layers, allowing you to move or rewrite words without disturbing the composition around them.

Lovart understands your style and keeps it consistent across iterations, formats, and projects. As contexts change, your voice remains unmistakably yours.