【摘要】 在当今科技赋能的文化传播过程中，三维内容生成工作已经成为了研究热点；得益于扩散模型与参数化三维表示模型的快速发展，基于二维扩散生成模型优化的三维内容生成技术已经出现众多令人惊叹的结果。其中，分数蒸馏采样技术则是利用二维扩散损失优化三维内容生成的重要环节。本文首先介绍扩散模型、三维表示方法和分数蒸馏采样等技术，然后汇总介绍基于分数蒸馏采样技术的三维内容生成代表性工作以及针对分数蒸馏采样技术不足的改进工作，此外进一步阐述领域相关的工作成果，并在最后对相关工作可能的改进方法和其他领域的应用作出展望。更多还原

【Abstract】 In the technology-enabled cultural communication process, 3D content generation has become a research hotpot. Thanks to the rapid development of diffusion models and parameterized 3D representation models, 3D content generation technology based on the optimization of 2D diffusion generation models has produced many amazing results, in which score distillation sampling is an important link in optimizing 3D content generation using 2D diffusion loss. In this paper at first the diffusion model, 3D representation methods and score distillation sampling were introduced. Then the representative works of 3D content generation based on score distillation sampling, and the improvement works for the shortcomings of score distillation sampling technology were summarized. Moreover, the results of the related works were further elaborated on. Finally the possible improvements of the works and prospects for applications in other fields was summarized.更多还原