pixelNeRF: pixelNeRF: Neural Radiance Fields from One or Few Images [CVPR 2021]
学习场景先验信息的NeRF
将输入图片也作为新视角合成的推断条件,而非仅仅是新视角的位姿,解决使用cost volume只能针对单个场景且需要优化的问题。
Depth Field Networks for Generalizable Multi-view Scene Representation [ECCV22]
这篇文章把相机位姿傅里叶变换之后作为网络的输入,可以参考。
Perceiver IO 解决Transformer: the quadratic scaling of self-attention with input size.
GAUDI: A Neural Architect for Immersive 3D Scene Generation [Arxiv]
[Porjecte page] 采样场景和轨迹两个latents来生成一段漫游场景的视频。
PREVIOUSPaper Reading
NEXTPaper Reading