search

ranahanocka / MeshCNN

Convolutional Neural Network for 3D meshes in PyTorch
MIT License
1.61k stars 320 forks source link

How to get the Mesh output? #28

Closed sawadika213 closed 5 years ago

sawadika213 commented 5 years ago

Thank you for your great work! I run your code on my local machine.But I have a little question. I use a Cube Mesh and operate a MeshConv on it , I can get the output features, but how can I get the mesh result?

TIM截图20190918172235

and output

image

ranahanocka commented 5 years ago

Not sure what you mean.

Anyway, please start by taking a look at the examples scripts (e.g., for cubes, see the train, test, and viewer).

sawadika213 commented 5 years ago

Sorry，

I mean I do a MeshConvlution of a mesh(cube mesh, not dataset...) .

I just get the features but I don't how to get a mesh using output features.

Thanks for your reply , I'm new about it.

ranahanocka commented 5 years ago

Hi @sawadika213 ,

Currently the operation you have done is take some mesh (presumably a cube), and extract the simple geometric features per edge (5-dimensional feature vector). Then you created a mesh convolution layer, and applied it on the 5-dimensional feature vector and brought it to 2-dimensions. So, all you did was change the edge features. The mesh itself (the topology, the connectivity) stayed the same. If you use meshpooling, then the mesh will change.

I suggest going through the examples in this repo, I have tools for visualizing the results, and I think it will be more clear.

sawadika213 commented 5 years ago

Exactly, I feel the MeshConv is very different from the convlution on image.I understand it.

Thank you reply!!