You are viewing the course site for a past offering of this course. The current offering may be found here.
Lecture 8: Meshes and Geometry Processing (75)

I don't see why this is beneficial exactly. In a object with regions that vary in detail, isn't having non-uniform triangles good? The uniform face is noticeably worse in the eyes, nose, and lips, but adds a bunch of complexity to the cheeks where it isn't really needed.


@leoadberg I think computationally, it's easier to deal with uniform objects. With a loss of uniformity, the ability to optimize probably decreases.


In what way do we decide whether to sacrifice the ability to optimize vs the detail we get with the non-uniform triangles? In most production level scenarios won't we want to get the most detail possible?


In the introduction of this paper, it seems that the motivation for remeshing is to make the mesh more suitable for later processes in the pipeline, since the source for the original mesh (e.g. a laser scan of an object) might produce a highly irregular mesh.


I feel like the non regular triangle meshes form more details of the shape than regular triangle meshes.


I suppose even though non-uniformity would model it better, in practice the trade offs in computation using uniform triangles are just probably more worth it. Its definitely simpler to work with uniformity across all triangles.


Adding onto what pervious comments have said, isn't the image on the right more accurate? I understand that there were rules set out for a 'good' mesh, but in an imperfect world isn't accuracy going to be found in more nuanced images (e.g. ones where all triangles are not uniform)?


It seems like there is more detail in the irregular image. So is this just a optimization tradeoff that we accept?


I believe that, yes, we are sacrificing the original detail for optimization. But since we are often required to do this at some step of the pipeline it is beneficial to do it here, early and under good practices.


I don't understand why having equally spaced triangles would help you calculate things faster, its the same number of numbers to work with.


there does seem to be more detail in the uniform image, but I think that's just because the areas where the triangles became very small are more densely sampled and the areas with larger triangles are less densely sampled. So really, there is more detail around the curved areas and less detail around the flatter areas on the irregular triangle rendering, but we don't perceive the less detailed area here because it's very low frequency in those flat areas anyway. The uniform shape and size make for more uniform sampling.


It would also probably end up looking much better after being upsampled, since the left image might end up really funky if we do some loop subdivision.

You must be enrolled in the course to comment