Title ECON: Explicit Clothed humans Obtained from Normals
Author Yuliang Xiu1 Jinlong Yang1 Xu Cao2 Dimitrios Tzionas3 Michael J. Black1
Conf/Jour CVPR
Year 2023
Project ECON: Explicit Clothed humans Optimized via Normal integration (xiuyuliang.cn)
Paper ECON: Explicit Clothed humans Obtained from Normals (readpaper.com)

image.png

姿态稳定(ICON在难的姿势下较好地重建)+灵活拓扑(ECON还可以较好地重建宽松的衣服)

缺陷:
image.png

Read more »

Title ICON: Implicit Clothed humans Obtained from Normals
Author Yuliang Xiu1, Jinlong Yang1, Dimitrios Tzionas1,2, Michael J. Black1
Conf/Jour CVPR
Year 2022
Project ICON (mpg.de)
Paper ICON: Implicit Clothed humans Obtained from Normals (readpaper.com)

image.png

CVPR 2022 | ICON: 提高三维数字人重建的姿势水平 - 知乎 (zhihu.com)

输入:

  • 经过分割的着衣人类的RGB图像
  • 从图像估计得到的SMPL身体
    • SMPL身体用于指导ICON的两个模块:一个推断着衣人类的详细表面法线(前视图和后视图),另一个推断一个具有可见性感知的隐式表面(占用场的等值表面)
    • 迭代反馈循环使用推断出的详细法线来优化SMPL

缺点:

  • 宽松的衣服无法重建
  • 依赖HPS估计出的SMPL body

image.png

Read more »

Title PIFuHD: Multi-Level Pixel-Aligned Implicit Function for High-Resolution 3D Human Digitization
Author Shunsuke Saito1,3 Tomas Simon2 Jason Saragih2 Hanbyul Joo3
Conf/Jour CVPR
Year 2020
Project PIFuHD: Multi-Level Pixel-Aligned Implicit Function for High-Resolution 3D Human Digitization (shunsukesaito.github.io)
Paper PIFuHD: Multi-Level Pixel-Aligned Implicit Function for High-Resolution 3D Human Digitization (readpaper.com)

image.png

Encoder: stacked hourglass network
MLP

  • Coarse L:(257, 1024, 512, 256, 128, 1)
  • Fine H:(272, 512, 256, 128, 1),将Coarse MLP的第四层输出$\Omega \in \mathbb{R}^{256}$作为输入
    表面法线网络:由9个残差块和4个下采样层组成
  • $\mathcal{L}_{N}=\mathcal{L}_{VGG}+\lambda_{l1}\mathcal{L}_{l1},$ 其中$L_{VGG}$为Johnson等人[17]提出的感知损失,$L_{l1}$为预测与真值法向之间的l1距离
Read more »

Title PIFu: Pixel-Aligned Implicit Function for High-Resolution Clothed Human Digitization
Author Shunsuke Saito1,2 Zeng Huang1,2 Ryota Natsume3 * Shigeo Morishima3 Angjoo Kanazawa4Hao Li1,2,5
Conf/Jour ICCV
Year 2019
Project PIFu: Pixel-Aligned Implicit Function for High-Resolution Clothed Human Digitization (shunsukesaito.github.io)
Paper PIFu: Pixel-Aligned Implicit Function for High-Resolution Clothed Human Digitization (readpaper.com)

image.png
表面重建网络:stacked hourglass
纹理推断网络:由残差块组成的architecture of CycleGAN
隐函数网络:MLP

Read more »

Title Recovering Fine Details for Neural Implicit Surface Reconstruction
Author Decai Chen1, Peng Zhang1,2, Ingo Feldmann1, Oliver Schreer1, and Peter Eisert1,3
Conf/Jour WACV
Year 2023
Project fraunhoferhhi/D-NeuS: Recovering Fine Details for Neural Implicit Surface Reconstruction (WACV2023) (github.com)
Paper Recovering Fine Details for Neural Implicit Surface Reconstruction (readpaper.com)

image.png

Idea:两个额外的损失函数

  • 几何偏差损失:鼓励隐式SDF场和体渲染的亮度场之间的几何一致性
  • 多视图特征一致性损失:多个观察视图在表面点处的特征一致
Read more »

Title Improving Neural Indoor Surface Reconstruction with Mask-Guided Adaptive Consistency Constraints
Author Xinyi Yu1, Liqin Lu1, Jintao Rong1, Guangkai Xu2,∗ and Linlin Ou1
Conf/Jour
Year 2023
Project
Paper Improving Neural Indoor Surface Reconstruction with Mask-Guided Adaptive Consistency Constraints (readpaper.com)

image.png

Idea:

  • 法向预测网络,法向量约束
  • 一致性约束(几何一致性和颜色一致性),通过虚拟视点实现
  • mask的计算方法,只计算有价值的光线
    Read more »

Title A Hierarchical Representation Network for Accurate and Detailed Face Reconstruction from In-The-Wild Images
Author Biwen Lei Jianqiang Ren Mengyang Feng Miaomiao Cui Xuansong Xie
Conf/Jour CVPR
Year 2023
Project HRN (younglbw.github.io)
Paper A Hierarchical Representation Network for Accurate and Detailed Face Reconstruction from In-The-Wild Images (readpaper.com)

image.png

缺点:

  • 需要3D 先验:每张图像的GT变形图和位移图
    Idea:
  • Contour-aware Loss. 新的轮廓感知损失算法,目的是拉动边缘的顶点以对齐面部轮廓
Read more »

Title Vid2Avatar: 3D Avatar Reconstruction from Videos in the Wild via Self-supervised Scene Decomposition
Author Chen Guo1, Tianjian Jiang1, Xu Chen1,2, Jie Song1, Otmar Hilliges1
Conf/Jour CVPR 2023
Year 2023
Project Vid2Avatar: 3D Avatar Reconstruction from Videos in the Wild via Self-supervised Scene Decomposition (moygcc.github.io)
Paper Vid2Avatar: 3D Avatar Reconstruction from Videos in the Wild via Self-supervised Scene Decomposition (readpaper.com)

image.png

Idea:$\mathcal{L}_\mathrm{dec}=\lambda_\mathrm{BCE}\mathcal{L}_\mathrm{BCE}+\lambda_\mathrm{sparse}\mathcal{L}_\mathrm{sparse}.$

  • 不透明度稀疏正则化$\mathcal{L}_{\mathrm{sparse}}^i=\frac1{|\mathcal{R}_{\mathrm{off}}^i|}\sum_{\mathbf{r}\in\mathcal{R}_{\mathrm{off}}^i}|\alpha^H(\mathbf{r})|.$惩罚与subject不相交的光线的非零光线不透明度
  • 自监督射线分类$\begin{aligned}\mathcal{L}_\mathrm{BCE}^i&=-\frac{1}{|\mathcal{R}^i|}\sum_{\mathrm{r}\in\mathcal{R}^i}(\alpha^H(\mathbf{r})\log(\alpha^H(\mathbf{r}))\\&+(1-\alpha^H(\mathbf{r}))\log(1-\alpha^H(\mathbf{r}))),\end{aligned}$鼓励包含完全透明或不透明光线的光线分布
    Read more »

Title High-fidelity 3D Human Digitization from Single 2K Resolution Images
Author Sang-Hun Han1, Min-Gyu Park2, Ju Hong Yoon2,Ju-Mi Kang2, Young-Jae Park1, and Hae-Gon Jeon1
Conf/Jour CVPR 2023 Highlight
Year 2023
Project High-fidelity 3D Human Digitization from Single 2K Resolution Images Project Page (sanghunhan92.github.io)
Paper High-fidelity 3D Human Digitization from Single 2K Resolution Images (readpaper.com)

image.png

可以看成一种估计深度图的方法
缺点:需要好的数据集

  • 需要提供法线图、mask、深度图(低分辨率+高分辨率)
  • 需要人体模型的关节点信息
  • 无法预测自遮挡部位
  • 对低分辨率重建效果不好
Read more »

NeRF-based重建方法之于前作监督的重建(新视图生成)方式,如MVS需要真实的深度图作监督,之前的包括生成式的方法需要三维模型的信息(PointCloud、Voxel、Mesh)作监督,NeRF-based方法构建了一种自监督的重建方式,从图像中重建物体只需要用图像作监督

NeRF将三维空间中所有点,通过MLP预测出对应的密度/SDF,是一种连续的方法(理论上,实际上由于计算机精度还是离散的)。至少在3D上不会由于离散方法(voxel),而出现很大的锯齿/aliasing

NeRF-based self-supervised 3D Reconstruction

  1. image and pose(COLMAP)
  2. NeRF(NeuS) or 3DGS(SuGaR)
    1. 损失函数(对比像素颜色、深度、法向量、SDF梯度累积<Eikonal term>Eikonal Equation and SDF - Lin’s site)
  3. PointCloud后处理,根据不同用途如3D打印、有限元仿真分析、游戏assets,有许多格式mesh/FEMode/AMs
Read more »

Title Flexible Isosurface Extraction for Gradient-Based Mesh Optimization
Author Shen, Tianchang and Munkberg, Jacob and Hasselgren, Jon and Yin, Kangxue and Wang, Zian and Chen, Wenzheng and Gojcic, Zan and Fidler, Sanja and Sharp, Nicholas and Gao, Jun
Conf/Jour ACM Trans. on Graph. (SIGGRAPH 2023)
Year 2023
Project Flexible Isosurface Extraction for Gradient-Based Mesh Optimization (FlexiCubes) (nvidia.com)
Paper Flexible Isosurface Extraction for Gradient-Based Mesh Optimization (nv-tlabs.github.io)

一种新的Marching Cube的方法
image.png

Read more »

Title BakedSDF: Meshing Neural SDFs for Real-Time View Synthesis
Author Lior Yariv and Peter Hedman and Christian Reiser and Dor Verbin and Pratul P. Srinivasan and Richard Szeliski and Jonathan T. Barron and Ben Mildenhall
Conf/Jour SIGGRAPH
Year 2023
Project BakedSDF
Paper BakedSDF: Meshing Neural SDFs for Real-Time View Synthesis (readpaper.com)

image.png

  • 对前景物体采用类似VolSDF的方法训练Modeling density with an SDF
  • 使用Marching Cube 的方法来提取高分辨率网格Baking a high-resolution mesh
  • Modeling view-dependent appearance,对baked后的高分辨率网格上顶点:采用漫反射颜色和球形高斯叶(前景3个波瓣,远处背景使用1个波瓣)
    • $\mathbf{C}=\mathbf{c}_{d}+\sum_{i=1}^{N}\mathbf{c}_{i}\exp\left(\lambda_{i}\left(\mu_{i}\cdot\mathbf{d}-1\right)\right).$
Read more »

Title Ref-NeuS: Ambiguity-Reduced Neural Implicit Surface Learning for Multi-View Reconstruction with Reflection
Author Wenhang Ge1 Tao Hu 2 Haoyu Zhao 1 Shu Liu 3 Ying-Cong Chen1,∗
Conf/Jour ICCV Oral
Year 2023
Project Ref-NeuS (g3956.github.io)
Paper Ref-NeuS: Ambiguity-Reduced Neural Implicit Surface Learning for Multi-View Reconstruction with Reflection (readpaper.com)
  • Anomaly Detection for Reflection Score + Visibility Identification for Reflection Score
  • Reflection Direction Dependent Radiance反射感知的光度损失
Read more »