zju3dv / ENeRF

SIGGRAPH Asia 2022: Code for "Efficient Neural Radiance Fields for Interactive Free-viewpoint Video"
https://zju3dv.github.io/enerf
Other
414 stars 28 forks source link

near_far插值范围限定 #14

Closed chky1997 closed 1 year ago

chky1997 commented 1 year ago

您好,目前代码利用了拟合的单人SMPL模型6890各顶点得到了3Dbbox,用来限定cost volume插值的深度(near_far),请问像project page里展示的多人情况和带背景的情况,是不用这个方法进行处理吗?

haotongl commented 1 year ago

Please refer to our supplementary material. In this complex scene, we layered the scene, established ENeRF separately, and used volume rendering to synthesize these layers. We used a large 3D bounding box composed of two people's SMPL points as the foreground layer, and the background layer was determined using the background point cloud. We recently released the ENeRF-Outdoor dataset, and we expect to release the relevant code before the Spring Festival. ———————————————————————————————————————————————————————————— 请参考我们的补充材料,arixv version paper Figure 6。我们在这种复杂的场景下,对场景进行了分层,分别建立了ENeRF,并且使用Volume rendering将这些层合成起来。我们使用两个人的SMPL点组成的大的3D bbox作为前景层,使用背景点云确定了背景层。我们最近公开了ENeRF-Outdoor dataset,预计春节前可以把相关code也release掉。

chky1997 commented 1 year ago

谢谢,期待!