HTDerekLiu/neuralSubdiv

Can't subdivide TOSCA shapes

antonognik opened this issue · 9 comments

antonognik commented

I'm trying to subdivide the TOSCA shapes (Non-rigid world) that were used in the paper for comparing the different methods.

For all shapes, except 3, I either get the error 'mesh has boundary' or 'edge non-manifold' in the SSP_decimation.m script.

HTDerekLiu commented

ohh we have a newer version of this that supports mesh with boundaries here https://github.com/HTDerekLiu/surface_multigrid_code/tree/main/09_random_subdiv_remesh.

You could try this out to see whether it works :)

antonognik commented

Thanks a lot for the reply :)

I tried the newer version and it works for meshes with boundaries. However, it doesn't work when the input mesh is not manifold.

What I'm trying to do is make some modifications in NeuralSubdivision and compare it with the rest of the results as seen in Table 2 of the paper:
image

I'm using the Non-rigid world meshes. Were these the ones used in the original paper?

Currently, from the centaur category only centaur1 and centaur2 are being decimated. The rest of the category are not manifold.

HTDerekLiu commented

Hmm I can't recall where I found a cleaned manifold version of the dataset. But I was able to locate the meshes I used in the original paper in my hard drive, see https://www.dropbox.com/scl/fi/rhpaq4y2vzthqxel9bh28/subdiv_results.zip?rlkey=n4eq5ti4tukdoy4mxgenj3wq1&dl=0.

Hope this helps!

antonognik commented

This certainly helps!

The only thing that I'm missing are the subdivided meshes used for the training of the network. (So I can train the modified one).

If you could find those I'd be really grateful

antonognik commented

Hey :)

Thanks a lot for the help so far!

I have one final question.

Do you have the other categories' (David, horse) shapes you trained the network with? (like the ones above)

therateful commented

hello, could you please share a Tosca dataset with me? The official website seems to have canceled the resource download

HTDerekLiu commented

Hi, you can download TOSCA from Vova's website http://www.vovakim.com/projects/CorrsBlended/doc_data.html

therateful commented

thank you very much