geometry-processing-parameterization icon indicating copy to clipboard operation
geometry-processing-parameterization copied to clipboard

Result u,v parameterization flipped

Open jnyao opened this issue 6 years ago • 3 comments

When I implement the Tutte method, it seems that the resulting flattened mesh gets flipped with its normal pointing downwards. Other than that, the distortion looks similar to the handout. Could you tell me what might be the reason for the flipping? Thank you!

jnyao avatar Oct 28 '18 21:10 jnyao

I think it might be because for the car, it has multiple boundaries as computed by igl::boundary_loop so there's no guarentee that the resulting parameterization is orientation preserving (since we only have proved that it's orientation preserving when the mesh has a single boundary only.)

AdrianShe avatar Oct 30 '18 02:10 AdrianShe

I think flipping is defined with respect to the other parametrized points. If all the normals of the parametrized points are on the same side then that is okay, right? That is, if n_i(uv) dot n_j(uv) = n_i(R3) dot n_j(R3) for all points i, j, then it is not a flipped parametrization (where n_i(uv) denotes the normal at point i in the parametrized space, and n_i(R3) denotes the corr normal in R^3).

vinbhaskara avatar Oct 30 '18 02:10 vinbhaskara

Since the optimization only cares about preserving angles, it would make sense that the map has a random up/down orientation. Right? Or am I missing something? Do we need to flip it manually after or can I leave it as is? I could check the normal of some triangle and flip the map as needed. Or is there a better way to avoid the flip altogether?

patrick-boc avatar Nov 02 '20 20:11 patrick-boc