search

nmwsharp / diffusion-net

Pytorch implementation of DiffusionNet for fast and robust learning on 3D surfaces like meshes or point clouds.
https://arxiv.org/abs/2012.00888
MIT License
398 stars 50 forks source link

Mesh Correspondence Example #5

Open mjkmoynihan opened 2 years ago

mjkmoynihan commented 2 years ago

Hi folks,

Thanks for sharing the code for this paper, it's a great read and really highlights some of the limitations of the state of the art. Would it be possible to share the code example for mesh correspondence?

I see that in the paper you follow the experimental set up of Donati et. al for this task. Does this refer only to the training strategy or does it mean you also adapt their functional map correspondence architecture after acquiring features?

Thanks!

nmwsharp commented 2 years ago

Hi!

We used the full functional map setup for the correspondence results in Sec 5.3. That is, we used a DiffusionNet to extract features, then those features were used as inputs to functional map correspondence, differentiating all the way back through the DiffusionNet for training.

We're working on getting that example up ASAP, it is the most involved one :) A closely related codebase can be found here, if you're looking for code: https://github.com/pvnieo/GeomFmaps_pytorch