Seems like the fact that a point cloud is hard to process/simulate outweighs the benefit of it being an easy way to represent large datasets (people probably want to process large datasets). Am I right to assume that the only reason you would ever use a point cloud is if you have no other reasonable way to model your geometry?
hesper
Is it also to use point clouds when we care about the entire volume in 3-d and not just the surface? Somehow I feel meshes are more about the surfaces?
hii
How do the costs of explicit/implicit representations compare?
motoole2
@treetrunk @hesper Points are really good at modeling.. points (duh!). If we want a geometric model for either a surface or volume, there are better representations. However, there are scenarios in which you are forced to work with point clouds. For example, when using a 3D scanner to capture real-world geometry, the 3D scanner usually outputs 3D points because these devices sample measurements discretely.
@hii Your question on the tradeoffs between explicit/implicit representations is tough to answer. In short, as stated earlier in this lecture, (1) implicit geometry can be used to determine how far a point x is from a surface (distance functions) or whether a point is inside or outside the surface, but converting this representation of geometry to a set of vertices is hard (higher computational cost to evaluate this); (2) explicit geometry provides ways to sample vertices on the surface directly, but determining whether vertices are inside or outside the object is hard (higher cost).
byungjul
I am really interested in technologies that convert point cloud to mesh! like (https://metastage.com/).
I know it is really advanced techniques and difficult to understand the whole pipelines. However, I just want to know what kind of fundamental concepts related to this process?
Seems like the fact that a point cloud is hard to process/simulate outweighs the benefit of it being an easy way to represent large datasets (people probably want to process large datasets). Am I right to assume that the only reason you would ever use a point cloud is if you have no other reasonable way to model your geometry?
Is it also to use point clouds when we care about the entire volume in 3-d and not just the surface? Somehow I feel meshes are more about the surfaces?
How do the costs of explicit/implicit representations compare?
@treetrunk @hesper Points are really good at modeling.. points (duh!). If we want a geometric model for either a surface or volume, there are better representations. However, there are scenarios in which you are forced to work with point clouds. For example, when using a 3D scanner to capture real-world geometry, the 3D scanner usually outputs 3D points because these devices sample measurements discretely.
@hii Your question on the tradeoffs between explicit/implicit representations is tough to answer. In short, as stated earlier in this lecture, (1) implicit geometry can be used to determine how far a point
xis from a surface (distance functions) or whether a point is inside or outside the surface, but converting this representation of geometry to a set of vertices is hard (higher computational cost to evaluate this); (2) explicit geometry provides ways to sample vertices on the surface directly, but determining whether vertices are inside or outside the object is hard (higher cost).I am really interested in technologies that convert point cloud to mesh! like (https://metastage.com/). I know it is really advanced techniques and difficult to understand the whole pipelines. However, I just want to know what kind of fundamental concepts related to this process?
Thanks!