JiayuYANG / Self-supervised-CVP-MVSNet

Self-supervised Learning of Depth Inference for Multi-view Stereo (CVPR 2021)
86 stars 12 forks source link

About the number of the point clouds and the time of mesh reconstruction for a single. #7

Open zhuixunforever opened 2 years ago

zhuixunforever commented 2 years ago

Hi, great work!

I want to know the number of 3D points of pseudo point clouds and the time of mesh reconstruction during your self-training.

XYZ-qiyh commented 2 years ago

Hello, @zhuixunforever The number of 3D points may depend on the resolution of the depth map, the parameter of depth filtering and depth fusion. In addition, the generation of the pseudo label (depth fusion, surface recon and depth rendering) cloud be viewed as offline processing.