search

facebookresearch / DeepSDF

Learning Continuous Signed Distance Functions for Shape Representation
MIT License
1.36k stars 255 forks source link

point cloud completion #67

Open yjcaimeow opened 3 years ago

yjcaimeow commented 3 years ago

Hi @tschmidt23 ,

Here is a question about how to get the empty-space (between and camera) as metioned below Table4 in the main paper?

Do the point cloud completion need to know the camera position?

How to get the camera position? I see the shapenet do not provide.

Best, YIngjie

hyperzy commented 3 years ago

Looks like all the objects in the same category have similar poses to the camera, as mentioned by the author, i.e.,

DeepSDF currently assumes models are in a canonical pose