geodist

[Zhangyanghao520]

Hello, could you please provide the files in the deogist folder?

[nicolasdonati]

Hi, Thanks for the interest :) I cannot upload geodist files because they are too heavy. I used matlab code from here https://github.com/llorz/SGA19_zoomOut/blob/master/utils/%2BMESH/geodesics_to_all.m but I will try to upload python code for this in coming weeks.

[zzifer]

Hi,May I ask when the file deogist is uploaded

[nicolasdonati]

Hi, A python code to generate geodesic distance matrices is available here: https://github.com/RobinMagnet/pyFM/blob/6d3e59a5155e040502115df6635dfa86f7b47c42/pyFM/mesh/geometry.py#L280 It is (partially) based on Mick Sharp's C++ implementation, which also has python bindings: https://github.com/nmwsharp/potpourri3d

[zzifer]

think you for you reply,and i have a problem.Because scape training set is not provided in your code,i downloaded dataset scape from https://graphics.soe.ucsc.edu/private/data/SCAPE/.But an error occurred at runtime. The error is: Process finished with exit code 139(interrupted by signal 11:SIGSEGV) The reason for the error might be because I used the wrong training set.Could you check it for me。think you

[zzifer]

The problem is solved after I convert the ply file to the off file. But a new problem has arisen. IndexError:index 20716 is out of bounds for axis 0 with size 1

• .-sad....55555555555555

[nicolasdonati]

I need more of the error to solve it. The data I used is this: https://nuage.lix.polytechnique.fr/index.php/s/LJFXrsTG22wYCXx

[zzifer]

i love you~mua mua mua~ I can't run code on my current computer. I'll contact you tomorrow

[zzifer]

hi,Thanks for the dataset you provided. Now the program can run correctly.I guess the previous problem was that there was no correspondence between the corres file and the shapes_train file. Here are some of my questions： 1.What do the a and r in the datasets scape_r and scape_a mean? 2.There seems to be a difference between the scape data set I got from https://ai.stanford.edu/~drago/Projects/scape/scape.html and the dataset you use. May I ask how these data have been processed? 3.I got the faust dataset from http://faust.is.tue.mpg.de/, but there are only ply, png and txt files in it. Do I need to convert txt to vts? Or what settings need to be modified?

[zzifer]

[nicolasdonati]

Ok so for FAUST and SCAPE original shapes, the ground-truth correspondence is 1 to 1 (basically the identity map). For the shapes I provided: FAUST_r and SCAPE_r (r is for re-meshed, as in here), the re-meshing shuffles the vertices so we keep track of this in vts files, which we use to recover the ground-truth map. The take-home message is that for original shapes, you can replace say X.vts by range(0, X.verts.shape[0]).

The _a datasets have been re-meshed by another method (it is detailed in the paper, I think supplementary). _a stands for anisotropic as the purpose of these re-meshed shapes is to create difficult non-regular triangulation to show that some method overfit to triangulation information.

For the third question, I think it is answered in my first point :) I hope that helps !

[zzifer]

Thank you for your enthusiastic answer these days.I will contact you if I have any questions later.I wish you more publications at top conferences！！！