# 计图开源：基于动态编码云的深度隐式函数学习

2022/06/09

## 计图开源：基于动态编码云的深度隐式函数学习

Part1

Part2

Part3

1）量化比较

Thingi32数据集上的拟合实验中，本文采用Chamfer Distance(CD)和Intersection over Union(IoU)作为量化指标来度量精度和重建质量，同时用潜在编码的数量和网络参数量度量方法的效率和存储、计算代价。表1为此实验的量化结果，从表中可以看到本文的方法优于其它方法，在同等的参数量和潜在编码数量的条件下达到了更高的精度，在近似的精度下本文的方法使用更少的潜在编码，却具有更高的效率。

ShapeNet数据集上的重建实验中，本文采用Chamfer Distance(CD)和F-Score作为量化指标，表2为此实验的量化结果，可以看出，该方法在CD和F-Score两个指标上都超越了已有的方法，证明了DCC-DIF学习形状先验和泛化的能力。

2）可视化比较

图4.ShapeNet可视化结果

https://lity20.github.io/DCCDIF_project_page

https://github.com/lity20/DCCDIF-jittor

1. Tianyang Li, Xin Wen, Yu-Shen Liu, Hua Su, and Zhizhong Han, Learning Deep Implicit Functions for 3D Shapes with Dynamic Code Clouds, IEEE/CVF CVPR, 2022.

2. Baorui Ma, Yu-Shen Liu, Matthias Zwicker, Zhizhong Han, Surface Reconstruction from Point Clouds by Learning Predictive Context Priors, IEEE/CVF CVPR, 2022.

3. Baorui Ma, Yu-Shen Liu, Zhizhong Han, Reconstructing Surfaces for Sparse Point Clouds with On-Surface Priors, IEEE/CVF CVPR, 2022.

4. Baorui Ma, Zhizhong Han, Yu-Shen Liu, Matthias Zwicker, Neural-Pull: Learning Signed Distance Functions from Point Clouds by Learning to Pull Space onto Surfaces, ICML, 2021.

5. Zhang Chen, Yinda Zhang, Kyle Genova, Sean Fanello, Sofien Bouaziz, Christian Hane, Ruofei Du, Cem Keskin, Thomas Funkhouser, and Danhang Tang, Multiresolution Deep Implicit Functions for 3D Shape Representation, IEEE/CVF ICCV, 2021, 13087-13096.

6. Towaki Takikawa, Joey Litalien, K. Yin, Karsten Kreis, Charles T. Loop, Derek Nowrouzezahrai, Alec Jacobson, Morgan McGuire, and Sanja Fidler, Neural Geometric Level of Detail: Real-time Rendering with Implicit 3D Shapes, IEEE/CVF CVPR, 2021, 11358-11367.

GGC往期回顾