You are viewing the course site for a past offering of this course. The current offering may be found here.
Lecture 27: Inverse Graphics/Research In Angjoo's Group (72)
ZizhengTai

It seems to me the distance-based NNs don't necessarily capture the actual 3D shape of the object since it only cares about the closest point to the viewer. If the network takes in the viewing direction as a parameter, then it does need to know the full 3D shape of the object. Occupancy-based networks apparently seem to capture the full 3D shape of the object.

You must be enrolled in the course to comment