{"title":"DreamMapping:通过变异分布映射实现高保真文本到 3D 的生成","authors":"Zeyu Cai, Duotun Wang, Yixun Liang, Zhijing Shao, Ying-Cong Chen, Xiaohang Zhan, Zeyu Wang","doi":"arxiv-2409.05099","DOIUrl":null,"url":null,"abstract":"Score Distillation Sampling (SDS) has emerged as a prevalent technique for\ntext-to-3D generation, enabling 3D content creation by distilling\nview-dependent information from text-to-2D guidance. However, they frequently\nexhibit shortcomings such as over-saturated color and excess smoothness. In\nthis paper, we conduct a thorough analysis of SDS and refine its formulation,\nfinding that the core design is to model the distribution of rendered images.\nFollowing this insight, we introduce a novel strategy called Variational\nDistribution Mapping (VDM), which expedites the distribution modeling process\nby regarding the rendered images as instances of degradation from\ndiffusion-based generation. This special design enables the efficient training\nof variational distribution by skipping the calculations of the Jacobians in\nthe diffusion U-Net. We also introduce timestep-dependent Distribution\nCoefficient Annealing (DCA) to further improve distilling precision. Leveraging\nVDM and DCA, we use Gaussian Splatting as the 3D representation and build a\ntext-to-3D generation framework. Extensive experiments and evaluations\ndemonstrate the capability of VDM and DCA to generate high-fidelity and\nrealistic assets with optimization efficiency.","PeriodicalId":501174,"journal":{"name":"arXiv - CS - Graphics","volume":"284 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"DreamMapping: High-Fidelity Text-to-3D Generation via Variational Distribution Mapping\",\"authors\":\"Zeyu Cai, Duotun Wang, Yixun Liang, Zhijing Shao, Ying-Cong Chen, Xiaohang Zhan, Zeyu Wang\",\"doi\":\"arxiv-2409.05099\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Score Distillation Sampling (SDS) has emerged as a prevalent technique for\\ntext-to-3D generation, enabling 3D content creation by distilling\\nview-dependent information from text-to-2D guidance. However, they frequently\\nexhibit shortcomings such as over-saturated color and excess smoothness. In\\nthis paper, we conduct a thorough analysis of SDS and refine its formulation,\\nfinding that the core design is to model the distribution of rendered images.\\nFollowing this insight, we introduce a novel strategy called Variational\\nDistribution Mapping (VDM), which expedites the distribution modeling process\\nby regarding the rendered images as instances of degradation from\\ndiffusion-based generation. This special design enables the efficient training\\nof variational distribution by skipping the calculations of the Jacobians in\\nthe diffusion U-Net. We also introduce timestep-dependent Distribution\\nCoefficient Annealing (DCA) to further improve distilling precision. Leveraging\\nVDM and DCA, we use Gaussian Splatting as the 3D representation and build a\\ntext-to-3D generation framework. Extensive experiments and evaluations\\ndemonstrate the capability of VDM and DCA to generate high-fidelity and\\nrealistic assets with optimization efficiency.\",\"PeriodicalId\":501174,\"journal\":{\"name\":\"arXiv - CS - Graphics\",\"volume\":\"284 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-09-08\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - CS - Graphics\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2409.05099\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - CS - Graphics","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.05099","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
DreamMapping: High-Fidelity Text-to-3D Generation via Variational Distribution Mapping
Score Distillation Sampling (SDS) has emerged as a prevalent technique for
text-to-3D generation, enabling 3D content creation by distilling
view-dependent information from text-to-2D guidance. However, they frequently
exhibit shortcomings such as over-saturated color and excess smoothness. In
this paper, we conduct a thorough analysis of SDS and refine its formulation,
finding that the core design is to model the distribution of rendered images.
Following this insight, we introduce a novel strategy called Variational
Distribution Mapping (VDM), which expedites the distribution modeling process
by regarding the rendered images as instances of degradation from
diffusion-based generation. This special design enables the efficient training
of variational distribution by skipping the calculations of the Jacobians in
the diffusion U-Net. We also introduce timestep-dependent Distribution
Coefficient Annealing (DCA) to further improve distilling precision. Leveraging
VDM and DCA, we use Gaussian Splatting as the 3D representation and build a
text-to-3D generation framework. Extensive experiments and evaluations
demonstrate the capability of VDM and DCA to generate high-fidelity and
realistic assets with optimization efficiency.