Title NeRFactor: Neural Factorization of Shape and Reflectance Under an Unknown Illumination
Author Xiuming Zhang    Pratul P. Srinivasan    Boyang Deng    Paul Debevec    William T. Freeman    Jonathan T. Barron
Conf/Jour TOG 2021 (Proc. SIGGRAPH Asia)
Year 2021
Project NeRFactor (xiuming.info)
Paper NeRFactor: Neural Factorization of Shape and Reflectance Under an Unknown Illumination (readpaper.com)

image.png

贡献:

  • NeRFactor在未知光照条件下从图像中恢复物体形状和反射率
    • 在没有任何监督的情况下,仅使用重渲染损失、简单的平滑先验和从真实世界BRDF测量中学习的数据驱动的BRDF先验,恢复了表面法线、光能见度、反照率和双向反射分布函数(BRDFs)的3D神经场
  • 通过NeRFactor的分解,我们可以用点光或光探针图像重新照亮物体,从任意视点渲染图像,甚至编辑物体的反照率和BRDF
Read more »

Title Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields
Author Dor Verbin and Peter Hedman and Ben Mildenhall and Todd Zickler and Jonathan T. Barron and Pratul P. Srinivasan
Conf/Jour CVPR 2022 (Oral Presentation, Best Student Paper Honorable Mention)
Year 2022
Project Ref-NeRF (dorverbin.github.io)
Paper Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields (readpaper.com)

image.png

贡献:

  • 借鉴Mip-NeRF的IPE,提出一种新的IDE来编码方向向量
  • 表面法向通过Spatial MLP来预测,并通过$\mathcal{R}_{\mathrm{p}}=\sum_{i}w_{i}|\hat{\mathbf{n}}_{i}-\hat{\mathbf{n}}_{i}^{\prime}|^{2},$来正则化使得预测得到的法向量和进一步计算的反射更加平滑
    • 这些MLP预测的法线往往比梯度密度法线更平滑
    • $\hat{\mathbf{n}}(\mathbf{x})=-\frac{\nabla\tau(\mathbf{x})}{|\nabla\tau(\mathbf{x})|}.$Eq.3
  • 计算反射光的新渲染方式$\mathbf{c}=\gamma(\mathbf{c}_d+\mathbf{s}\odot\mathbf{c}_s),$
    • $\hat{\mathbf{\omega}}_r=2(\hat{\mathbf{\omega}}_o\cdot\hat{\mathbf{n}})\hat{\mathbf{n}}-\hat{\mathbf{\omega}}_o,$ Eq.4
    • $L_{\mathrm{out}}(\hat{\mathbf{\omega}}_{o})\propto\int L_{\mathrm{in}}(\hat{\mathbf{\omega}}_{i})p(\hat{\mathbf{\omega}}_{r}\cdot\hat{\mathbf{\omega}}_{i})d\hat{\mathbf{\omega}}_{i}=F(\hat{\mathbf{\omega}}_{r}).$ 借鉴此BRDF,提出的Direction MLP 得出$c_s$
    • 漫反射颜色$c_d$通过Spatial MLP预测得到
    • s是高光色调
    • 将空间MLP输出的瓶颈向量b传递到Direction MLP中,这样反射的亮度就可以随着3D位置的变化而变化。
  • $\mathcal{R}_{\mathrm{o}}=\sum_{i}w_{i}\max(0,\hat{\mathbf{n}}_{i}^{\prime}\cdot\hat{\mathbf{d}})^{2}.$ 正则化项惩罚朝向远离相机的法线

局限:

  • 编码导致的速度慢,和Spatial MLP的loss反向传播速度比Mip-NeRF慢
  • 没有明确地模拟相互反射或非远距离照明
    • 忽略互反射和自遮挡等现象
Read more »

Title ShadowNeuS: Neural SDF Reconstruction by Shadow Ray Supervision
Author Jingwang Ling and Zhibo Wang and Feng Xu
Conf/Jour CVPR
Year 2023
Project ShadowNeuS (gerwang.github.io)
Paper ShadowNeuS: Neural SDF Reconstruction by Shadow Ray Supervision (readpaper.com)

image.png
方法:假设场景不发光,且忽略相互反射

  • 从二值阴影图像中获得可见表面的入射亮度,然后处理更复杂的RGB图像
  • 入射光辐射$C_\mathrm{in}(x,l)=L\prod_{i=1}^N(1-\alpha_i)$, 从单视图、多光源中重建出3D shape
    • $\mathcal{L}_\mathrm{shadow}=|\widehat{C}_\mathrm{in}-I_\mathrm{s}|_1.$
  • 出射光辐射$C(x,-\mathbf{v})=(\rho_d+\rho_s)C_{\mathrm{in}}(x,l)(l\cdot\mathbf{n})$
    • $\mathcal{L}_\mathrm{rgb}=|\widehat{C}-I_\mathrm{r}|_1$

表现:outperforms the SOTAs in single-view reconstruction, and it has the power to reconstruct scene geometries out of the camera’s line of sight.

Read more »

使用$\hat{C}(\mathbf{r})=\sum_{i=1}^No(\mathbf{x}_i)\prod_{j<i}\bigl(1-o(\mathbf{x}_j)\bigr)c(\mathbf{x}_i,\mathbf{d})$ 占据o来代替NeRF中的$\alpha$
将VR与SR结合起来,首先根据占据场获取表面的点$t_s$,然后在$t_s$的一个区间内均匀采样点来进行颜色场的优化(如果光线没有穿过物体,则使用分层采样)

image.png|700

Read more »

Title PAniC-3D: Stylized Single-view 3D Reconstruction from Portraits of Anime Characters
Author Chen, Shuhong and Zhang, Kevin and Shi, Yichun and Wang, Heng and Zhu, Yiheng and Song, Guoxian and An, Sizhe and Kristjansson, Janus and Yang, Xiao and Matthias Zwicker
Conf/Jour CVPR
Year 2023
Project ShuhongChen/panic3d-anime-reconstruction: CVPR 2023: PAniC-3D Stylized Single-view 3D Reconstruction from Portraits of Anime Characters (github.com)
Paper PAniC-3D: Stylized Single-view 3D Reconstruction from Portraits of Anime Characters (readpaper.com)

基于EG3D无条件生成模型
PAniC-3D对比PixelNeRF、EG3D(+Img2stylegan or +PTI)、Pifu

image.png

Read more »

Title Few-View Object Reconstruction with Unknown Categories and Camera Poses
Author Hanwen Jiang Zhenyu Jiang Kristen Grauman Yuke Zhu
Conf/Jour ArXiv
Year 2022
Project FORGE (ut-austin-rpl.github.io)
Paper Few-View Object Reconstruction with Unknown Categories and Camera Poses (readpaper.com)

估计视图之间的相对相机姿态
image.png

贡献:

  • 2D提取voxel特征 —> 相机姿态估计 —> 特征共享+融合 —> MLP神经隐式重建 —> 渲染已有相机位姿的图片,并计算与gt之间的loss
  • 新的损失函数

image.png

Read more »

Zip-NeRF在抗混叠(包括NeRF从空间坐标到颜色和密度的学习映射的空间混叠,以及沿每条射线在线蒸馏过程中使用的损失函数的z-混叠)方面都取得了很好的效果,并且速度相比前作Mip-NeRF 360 提高了24X

mipNeRF 360+基于网格的模型(如Instant NGP)的技术

  • 错误率下降8~77%,并且比Mip-NeRF360提速了24X
  • 主要贡献:
    • 多采样
    • 预滤波

多采样:train左,test右

Image 1
Image 2
Read more »

Reference

[PDF] NeRD: Neural Reflectance Decomposition From Image Collections
[PDF] SAMURAI: Shape And Material from Unconstrained Real-world Arbitrary Image collections
[PDF] Relighting4D: Neural Relightable Human from Videos
[PDF] Neural 3D Scene Reconstruction with the Manhattan-world Assumption
[PDF] NeROIC: Neural Rendering of Objects from Online Image Collections

对金属反光材质的物体重建效果很好

imgae

提出了一种新的光表示方法,颜色由漫反射和镜面反射两部分组成,通过两个阶段的方法来实现

  • Stage1:使用集成方向编码来近似光积分,使用shadow MLP对直接光和间接光进行model,学习到了表面几何形状
  • Stage2:蒙特卡罗采样固定几何形状,重建更精确的表面BRDF和环境光
    • $\mathbf{c}_{\mathrm{diffuse}}=\frac{1}{N_{d}}\sum_{i}^{N_{d}}(1-m)\mathrm{a}L(\omega_{i}),$
    • $\mathbf{c}_{\mathrm{specular}}=\frac{1}{N_{s}}\sum_{i}^{N_{s}}\frac{FG(\omega_{0}\cdot\mathbf{h})}{(\mathbf{n}\cdot\mathbf{h})(\mathbf{n}\cdot\omega_{\mathbf{0}})}L(\omega_{i}),$
Read more »

Title Tri-MipRF: Tri-Mip Representation for Efficient Anti-Aliasing Neural Radiance Fields
Author Wenbo Hu Yuling Wang
Conf/Jour ICCV
Year 2023
Project Tri-MipRF: Tri-Mip Representation for Efficient Anti-Aliasing Neural Radiance Fields (wbhu.github.io)
Paper Tri-MipRF: Tri-Mip Representation for Efficient Anti-Aliasing Neural Radiance Fields (readpaper.com)

2023.7.26 SOTA
image.png|600

image.png

like TensoRF: Tensorial Radiance Fields (apchenstu.github.io)+ Mip-NeRF + NGP

  • 类似Mip-NeRF中的采样锥形方法,但是Tri-MipRF使用了与圆锥相切的采样球S=(x,r),代替Mip-NeRF中的多元高斯圆锥体

    • 采样球的半径通过像素圆盘半径$\dot r$(由像素大小in world Coordinate),焦距f和$t_i$确定
  • 类似TensoRF中的分解方法,将空间采样球分解到三个平面上,编码类似NGP中的HashGrid 使用2D平面来存取特征值,构建一个base level:$M^{L_{0}}$,通过downscaling来获得其他level的2D grid平面。

    • 通过base level中interest space的AABB求出$\ddot r$,并联合采样球半径r得到采样球在平面投影的level,根据此level和投影到平面上的二维坐标,在相邻两level $\mathcal{M}_{XY}^{\lfloor l\rfloor}$和$\mathcal{M}_{XY}^{\lceil l\rceil}$的2D grid中采用3线性插值得到采样球的特征值,最后三个分解平面的特征值cat起来作为MLP的一个输入
  • 一种更好的渲染视图方法:Hybrid Volume-Surface Rendering

    • 通过在密度场中marching cubes和网格抽取来获得代理网格,粗略确定相机原点到物体表面的距离
    • 对代理网格进行有效栅格化,以获得圆锥体中轴线表面上的命中点,然后我们在距圆锥体中轴线命中点∆t距离内均匀采样球体,这产生2∆t采样间隔。
    • 优点:可以减少需要采样点的数量,且不会影响渲染出来图片的质量
  • 优点:

    • fine-grained details in close-up views
    • and free of aliasing in distant views
    • 5 minute and smaller model parameters
  • 缺点:

    • 需要使用multi-view segmentation methods将In-the-wildIn数据集中感兴趣的物体提取出来
      • 即需要mask
Read more »

Title Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields
Author Jonathan T. Barron Ben Mildenhall Dor Verbin Pratul P. Srinivasan Peter Hedman
Conf/Jour CVPR 2022 (Oral Presentation)
Year 2022
Project mip-NeRF 360 (jonbarron.info)
Paper Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields (readpaper.com)
  • novel Kalman-like scene parameterization:将场景参数化,将单位球外背景采样的截头锥参数化到r=2的球体内,单位球内的截头锥不受影响

image.png

  • efficient proposal-based coarse-to-fine distillation framework:一个提议网络用来获取权重,用来进行精采样,再通过精采样的点根据NeRF 的MLP得到密度和颜色值

image.png

  • regularizer designed for mipNeRF ray intervals:可以有效消除floaters(体积密集空间中不相连的小区域)和背景塌陷(远处的表面被错误地模拟成靠近相机的密集内容的半透明云)

$\begin{gathered}\mathcal{L}_{\mathrm{dist}}(\mathbf{s},\mathbf{w}) =\sum_{i,j}w_iw_j\left|\frac{s_i+s_{i+1}}{2}-\frac{s_j+s_{j+1}}{2}\right| \+\frac13\sum_iw_i^2(s_{i+1}-s_i) \end{gathered}$

image.png

Read more »

Title Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields
Author Jonathan T. Barron and Ben Mildenhall and Matthew Tancik and Peter Hedman and Ricardo Martin-Brualla and Pratul P. Srinivasan
Conf/Jour 2021 IEEE/CVF International Conference on Computer Vision (ICCV)
Year 2021
Project google/mipnerf (github.com)
Paper Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields (readpaper.com)

image.png

  • 一种新的采样方式:锥体采样conical frustums截头圆锥体
  • 基于PE提出了IPE,可以平滑地编码空间体积的大小和形状
  • 将NeRF的粗精采样MLP合并为一个MLP

IPE:当锥体区域较宽(正态分布很宽)时,会将高频的信息积分为0;当区域较窄时,保持原来的PEncoding
ipe_anim_horiz.gif

image.png

Read more »

Title NeRF++: Analyzing and Improving Neural Radiance Fields
Author Kai Zhang, Gernot Riegler, Noah Snavely, Vladlen Koltun
Conf/Jour arXiv: Computer Vision and Pattern Recognition
Year 2020
Project Kai-46/nerfplusplus: improves over nerf in 360 capture of unbounded scenes (github.com)
Paper NeRF++: Analyzing and Improving Neural Radiance Fields. (readpaper.com)
arxiv.org/pdf/2010.07492

image.png

创新:一种前背景分离的方法
挑战:

  • First, the training and testing of NeRF and NeRF++ on a single large-scale scene is quite time-consuming and memory-intensive —> NGP解决了耗时
  • Second, small camera calibration errors may impede阻碍 photorealistic synthesis. Robust loss functions, such as the contextual loss (Mechrez et al., 2018), could be applied.
  • Third, photometric effects such as auto-exposure and vignetting渐晕 can also be taken into account to increase image fidelity. This line of investigation is related to the lighting changes addressed in the orthogonal work of Martin-Brualla et al. (2020).
Read more »

Title NeuDA: Neural Deformable Anchor for High-Fidelity Implicit Surface Reconstruction
Author Bowen Cai ,Jinchi Huang, Rongfei Jia ,Chengfei Lv, Huan Fu*
Conf/Jour CVPR
Year 2023
Project NeuDA (3d-front-future.github.io)
Paper NeuDA: Neural Deformable Anchor for High-Fidelity Implicit Surface Reconstruction (readpaper.com)

NeuDA变形后的grid距离Surface更近一些,即可以使采样点插值时更多依赖于表面,即渲染时也会更多地考虑到3D空间相邻的信息

创新:Deformable Anchors、HPE、$\mathcal{L}_{norm}$

  • 改进了NGP中的grid表示,8个顶点存储feature—>存储锚点位置,锚点位置经过PE后输入进SDF网络

image.png

image.png

Read more »

从仅2~3张的稀疏输入中重建表面

  • 首先,我们提出了一个多层几何推理框架,以粗到细的方式恢复表面。
  • 其次,我们采用了一种多尺度颜色混合方案,该方案联合评估局部和背景亮度一致性,以获得更可靠的颜色预测。
  • 第三,采用一致性感知的微调方案,控制遮挡和图像噪声引起的不一致区域,得到准确、干净的重建。

image.png

Read more »