wuvin
Always take risks!
Toggle navigation
wuvin
主页
实验室的搬砖生活
机器学习
公开的学术内容
公开的其他内容
About Me
归档
标签
友情链接
ZYQN
ihopenot
enigma_aw
hzwer
杨宗翰
3D 生成调研
2022-01-13 16:41:19
220
0
0
wuvin
# Voxel ### Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling  ### Octree Generating Networks: Efficient Convolutional Architectures for High-resolution 3D Outputs   ### Visual Object Networks: Image Generation with Disentangled 3D Representation  ### Improved Adversarial Systems for 3D Object Generation and Reconstruction  # Point Cloud ### Learning Representations and Generative Models for 3D Point Clouds   ### FoldingNet: Point Cloud Auto-encoder via Deep Grid Deformation   ### PointFlow: 3D Point Cloud Generation with Continuous Normalizing Flows   ### Shape Inpainting using 3D Generative Adversarial Network and Recurrent Convolutional Networks   ### Learning Efficient Point Cloud Generation for Dense 3D Object Reconstruction   # Mesh ### Neural 3D Mesh Renderer   ### SurfNet: Generating 3D shape surfaces using deep residual networks    ### BSP-Net: Generating Compact Meshes via Binary Space Partitioning     ### Pixel2Mesh++: Multi-View 3D Mesh Generation via Deformation   ### PolyGen: An Autoregressive Generative Model of 3D Meshes   # Other Important Papers ### Local Implicit Grid Representations for 3D Scenes (CVPR 2020)      * 这个相当于是 Grid + Occupancy Network。 所以是无法渲染出颜色，而且训练需要3D模型数据的。做的任务也只是 Memories 一个3D场景，本质上是 Nerf的竞争对手。 ### Local Deep Implicit Functions for 3D Shape   * 这个是用大量正态分布做的，不构成竞争。 # Conclusion * 除了苹果那篇室内场景生成【Unconstrained Scene Generation with Locally Conditioned Radiance Fields (ICCV 2021), 在之前的基于Nerf的生成模型的调研里】以外，目前为止没有任何一个工作做到了大型3D场景生成。 * 已有的基于 Voxel, Point Cloud, Mesh 的方法都没颜色。
上一篇:
Point-NeRF: Point-based Neural Radiance Fields
下一篇:
GRAF 后续
0
赞
220 人读过
新浪微博
微信
腾讯微博
QQ空间
人人网
提交评论
立即登录
, 发表评论.
没有帐号?
立即注册
0
条评论
More...
文档导航
wuvin
没有帐号? 立即注册