csm icon indicating copy to clipboard operation
csm copied to clipboard

pre-processed mesh template

Open anandbhattad opened this issue 5 years ago • 4 comments

Hi!

Can you please provide pre-processed mesh templates used?

best, Anand

anandbhattad avatar Jun 08 '20 14:06 anandbhattad

Hi, Please find the models here. https://umich.box.com/s/lik2xukm5qtn68lmqim60altus7k5o1g

Also, if you can access the mean_shape.mat file under cachedir/shapenet/<category> it also has the corresponding verts and faces which you can dump as an obj.

Best, Nilesh

nileshkulkarni avatar Jun 08 '20 22:06 nileshkulkarni

Thanks a lot, Nilesh! That was fast :)

Best, Anand

anandbhattad avatar Jun 08 '20 22:06 anandbhattad

Hi Nilesh,

I am trying to get a template mesh for another object. Simply running mesh deform doesn't seem to generate templates like the ones you shared. I realized reading from your ReadMe I need to complete the following 2, 3, 4 steps before doing mesh deform. Is this correct?

  1. Have a model of the category as an obj file.
  2. Convert the model to tsdf field.
  3. Run Marching cubes to get on model
  4. Model clean + close holes a. Merge Close vertices b. Quadratic Simplication c. Remove non-manifold edges
  5. Mesh-deform to create a parametrized mapping

How did you do 2,3,4? Do you have a code for it? I see here marching_cubes.py mentioned in the following line and I cannot locate marching_cubes.py file in the repo: https://github.com/nileshkulkarni/csm/blob/5923517269183f6fa73049e83c8db478c2899b07/csm/preprocess/parameterize/normalize/normalize_online_model.py#L44

Thanks for your help!

Best, Anand

anandbhattad avatar Jun 12 '20 17:06 anandbhattad

Hi Anand,

Step 2 can be done using this code. Alternately you can also run binvox on the mesh obj file, and then using that voxel volume as input to marching cubes.

Step 3 can be done using marching cubes from scipy

Step 4 completed in meshlab.

Best, Nilesh

nileshkulkarni avatar Jun 18 '20 04:06 nileshkulkarni