人工智能培训

搜索

人工智能论文:基于潜在空间分解的复合形状建模(Composite Shape Modeling via Latent Space Factorization)

[复制链接]
wjx003006 发表于 2019-1-11 10:30:31 | 显示全部楼层 |阅读模式
wjx003006 2019-1-11 10:30:31 220 0 显示全部楼层
人工智能论文:基于潜在空间分解的复合形状建模(Composite Shape Modeling via Latent Space Factorization)我们提出了一种新的神经网络架构,称为Decomposer-Composer,用于语义结构感知的3D形状建模。我们的方法利用基于自动编码器的管道,并产生一个新颖的分解形状嵌入空间,其中形状集合的语义结构转换为依赖于adata的子空间分解,并且形状组合和分解在嵌入坐标上变为简单的线性运算。我们进一步建议使用显式学习部件变形模块对形状装配进行建模,该模块利用3D空间变换器网络来实现网络内体积网格变形,并且允许我们对整个系统进行端到端训练。由此产生的网络允许我们执行部分级别的操作,这是现有方法无法实现的。我们广泛的消融研究,与基线方法和定性分析的比较证明了所提出方法的改进性能。
We present a novel neural network architecture, termed Decomposer-Composer,for semantic structure-aware 3D shape modeling.Our method utilizes anauto-encoder-based pipeline, and produces a novel factorized shape embeddingspace, where the semantic structure of the shape collection translates into adata-dependent sub-space factorization, and where shape composition anddecomposition become simple linear operations on the embedding coordinates.Wefurther propose to model shape assembly using an explicit learned partdeformation module, which utilizes a 3D spatial transformer network to performan in-network volumetric grid deformation, and which allows us to train thewhole system end-to-end.The resulting network allows us to perform part-levelshape manipulation, unattainable by existing approaches.Our extensive ablationstudy, comparison to baseline methods and qualitative analysis demonstrate theimproved performance of the proposed method.人工智能论文:基于潜在空间分解的复合形状建模(Composite Shape Modeling via Latent Space Factorization) bHT7Dvo5TVSAKdSd.jpg
URL地址:https://arxiv.org/abs/1901.02968     ----pdf下载地址:https://arxiv.org/pdf/1901.02968    ----人工智能论文:基于潜在空间分解的复合形状建模(Composite Shape Modeling via Latent Space Factorization)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

wjx003006当前离线
新手上路

查看:220 | 回复:0

快速回复 返回顶部 返回列表