| Title | DoubleField: Bridging the Neural Surface and Radiance Fields for High-fidelity Human Rendering |
|---|---|
| Author | Ruizhi Shao1, Hongwen Zhang1, He Zhang2, Yanpei Cao3, Tao Yu1, and Yebin Liu1 |
| Conf/Jour | CVPR |
| Year | 2022 |
| Project | DoubleField Project Page (liuyebin.com) |
| Paper | DoubleField: Bridging the Neural Surface and Radiance Fields for High-fidelity Human Rendering. (readpaper.com) |
| Title | Human as Points: Explicit Point-based 3D Human Reconstruction from Single-view RGB Images |
|---|---|
| Author | Yingzhi Tang, Qijian Zhang, Junhui Hou, and Yebin Liu |
| Conf/Jour | arXiv |
| Year | 2023 |
| Project | yztang4/HaP (github.com) |
| Paper | Human as Points—— Explicit Point-based 3D Human Reconstruction from Single-view RGB Images.pdf (readpaper.com) |
深度估计+SMPL 估计+Diffusion Model 精细化(PointNet++)
缺陷：依赖于深度估计方法和 SMPL 估计方法的精度
为了保护隐私不对人脸进行重建
| Title | CCD-3DR: Consistent Conditioning in Diffusion for Single-Image 3D Reconstruction |
|---|---|
| Author | Yan Di1, Chenyangguang Zhang2, Pengyuan Wang1, Guangyao Zhai1, Ruida Zhang2, Fabian Manhardt3, Benjamin Busam1, Xiangyang Ji2, and Federico Tombari1,3 |
| Conf/Jour | arXiv |
| Year | 2023 |
| Project | |
| Paper | CCD-3DR: Consistent Conditioning in Diffusion for Single-Image 3D Reconstruction (readpaper.com) |
No Code
| Title | Deep Marching Tetrahedra: a Hybrid Representation for High-Resolution 3D Shape Synthesis |
|---|---|
| Author | Tianchang Shen and Jun Gao and Kangxue Yin and Ming-Yu Liu and Sanja Fidler |
| Conf/Jour | NeurIPS |
| Year | 2021 |
| Project | Deep Marching Tetrahedra: a Hybrid Representation for High-Resolution 3D Shape Synthesis (nvidia.com) |
| Paper | Deep Marching Tetrahedra: a Hybrid Representation for High-Resolution 3D Shape Synthesis (readpaper.com) |
输入点云或低分辨率体素，提取特征后利用GAN网络，生成每个顶点的位置和SDF偏移值，得到优化后顶点的位置和SDF
结合显式与隐式表达的表示方法，利用MT，从隐式SDF中重建出显式mesh
| Title | Rodin: A Generative Model for Sculpting 3D Digital Avatars Using Diffusion |
|---|---|
| Author | Tengfei Wang1† Bo Zhang2 Ting Zhang2 Shuyang Gu2 Jianmin Bao2 |
| Conf/Jour | arXiv preprint |
| Year | 2022 |
| Project | RODIN Diffusion (microsoft.com) |
| Paper | Rodin: A Generative Model for Sculpting 3D Digital Avatars Using Diffusion (readpaper.com) |
| Title | TransHuman: A Transformer-based Human Representation for Generalizable Neural Human Rendering |
|---|---|
| Author | Xiao Pan1,2,∗, Zongxin Yang1, Jianxin Ma2, Chang Zhou2, Yi Yang1,† |
| Conf/Jour | ICCV |
| Year | 2023 |
| Project | TransHuman: A Transformer-based Human Representation for Generalizable Neural Human Rendering (pansanity666.github.io) |
| Paper | TransHuman: A Transformer-based Human Representation for Generalizable Neural Human Rendering (readpaper.com) |
PointCloud
Follow
| Paper | Model | Input | Parameter/Pnum | GPU |
|---|---|---|---|---|
| DiT-3D | Diffusion Transformers | Voxelized PC | ||
| PointFlow | AE flow-based | PointCloud | 1.61M | |
| FlowGAN | GAN flow-based | Single Image | N = 2500 | A40 45GB |
| BuilDiff | Diffusion models | Single Image | 1024 to 4096 | A40 45GB |
| CCD-3DR | CDPM | Single Image | 8192 | 3090Ti 24GB |
| SG-GAN | SG-GAN | Single Image | ||
| HaP | Diffusion+SMPL+DepthEstimation | Single Image | 10000 | 4x3090Ti |
| Title | BuilDiff: 3D Building Shape Generation using Single-Image Conditional Point Cloud Diffusion Models |
|---|---|
| Author | Wei, Yao and Vosselman, George and Yang, Michael Ying |
| Conf/Jour | ICCV |
| Year | 2023 |
| Project | weiyao1996/BuilDiff: BuilDiff: 3D Building Shape Generation using Single-Image Conditional Point Cloud Diffusion Models (github.com) |
| Paper | BuilDiff: 3D Building Shape Generation using Single-Image Conditional Point Cloud Diffusion Models (readpaper.com) |
| Title | SG-GAN: Fine Stereoscopic-Aware Generation for 3D Brain Point Cloud Up-sampling from a Single Image |
|---|---|
| Author | Bowen Hu, Baiying Lei, Shuqiang Wang, Senior Member, IEEE |
| Conf/Jour | arXiv |
| Year | 2023 |
| Project | |
| Paper | SG-GAN: Fine Stereoscopic-Aware Generation for 3D Brain Point Cloud Up-sampling from a Single Image (readpaper.com) |
Stereoscopic-aware graph generative adversarial network (SG-GAN)
| Title | Flow-based GAN for 3D Point Cloud Generation from a Single Image |
|---|---|
| Author | Yao Wei (University of Twente), George Vosselman (“University of Twente, the Netherlands”), Michael Ying Yang (University of Twente)* |
| Conf/Jour | BMVA |
| Year | 2022 |
| Project | Flow-based GAN for 3D Point Cloud Generation from a Single Image (mpg.de) |
| Paper | Flow-based GAN for 3D Point Cloud Generation from a Single Image (readpaper.com) |
| Title | DiT-3D: Exploring Plain Diffusion Transformers for 3D Shape Generation |
|---|---|
| Author | Shentong Mo 1, Enze Xie 2, Ruihang Chu 3, Lewei Yao 2,Lanqing Hong2, Matthias Nießner4, Zhenguo Li2 |
| Conf/Jour | arXiv |
| Year | 2023 |
| Project | DiT-3D: Exploring Plain Diffusion Transformers for 3D Shape Generation |
| Paper | DiT-3D: Exploring Plain Diffusion Transformers for 3D Shape Generation (readpaper.com) |
New 3D Diffusion Transformer Model, 在体素化的点云上运行 DDPM(Denoising diffusion probabilistic models) 的去噪过程
| Title | Challenging universal representation of deep models for 3D point cloud registration |
|---|---|
| Author | Bojani\’{c}, David and Bartol, Kristijan and Forest, Josep and Gumhold, Stefan and Petkovi\’{c}, Tomislav and Pribani\’{c}, Tomislav |
| Conf/Jour | BMVC |
| Year | 2022 |
| Project | DavidBoja/greedy-grid-search: [BMVC 2022 workshop] Greedy Grid Search: A 3D Registration Baseline (github.com) |
| Paper | Challenging the Universal Representation of Deep Models for 3D Point Cloud Registration (readpaper.com) |
按步长穷举法，粗配准，需要根据ICP进行精配准
| Title | Geometric Transformer for Fast and Robust Point Cloud Registration |
|---|---|
| Author | Zheng Qin1 Hao Yu2 Changjian Wang1 Yulan Guo1,3 Yuxing Peng1 Kai Xu1* |
| Conf/Jour | CVPR |
| Year | 2022 |
| Project | qinzheng93/GeoTransformer: [CVPR2022] Geometric Transformer for Fast and Robust Point Cloud Registration (github.com) |
| Paper | Geometric Transformer for Fast and Robust Point Cloud Registration (readpaper.com) |
pairwise registration models, only Ubuntu
| Title | Robust Multiview Point Cloud Registration with Reliable Pose Graph Initialization and History Reweighting |
|---|---|
| Author | Haiping Wang and Yuan Liu and Zhen Dong and Yulan Guo and Yu-Shen Liu and Wenping Wang and Bisheng Yang |
| Conf/Jour | CVPR |
| Year | 2023 |
| Project | WHU-USI3DV/SGHR: [CVPR 2023] Robust Multiview Point Cloud Registration with Reliable Pose Graph Initialization and History Reweighting (github.com) |
| Paper | Robust Multiview Point Cloud Registration with Reliable Pose Graph Initialization and History Reweighting (readpaper.com) |
Issue:
How should I train my dataset? · Issue #4 · WHU-USI3DV/SGHR (github.com)
I think several point clouds of a single statue is not enough for training deep descriptors. I suggest to directly use pairwise registration models such as Geotrainsformer pre-trained on object-level datasets such as ModelNet40 to solve the pairwise registrations.
And adopt SGHR’s transformation synchronization section to solve the global consistent scan poses.
Math basic of 3D Reconstruction