search

autonomousvision / occupancy_networks

This repository contains the code for the paper "Occupancy Networks - Learning 3D Reconstruction in Function Space"
https://avg.is.tuebingen.mpg.de/publications/occupancy-networks
MIT License
1.5k stars 292 forks source link

wateritght meshes #15

Open JeremyFisher opened 5 years ago

JeremyFisher commented 5 years ago

Hey there,

congrats on a great paper!

I saw that there are no watertight meshes in your "preprocessed data (73.4 GB)".

Would it be possible for you to upload the preprocessed, ie watertight, meshes as well?

xiayizhan2017 commented 5 years ago

@JeremyFisher Hi, can you share the data you downloaded? I always failed to download the link provided by the author. Thank you in advance!

LMescheder commented 5 years ago

@JeremyFisher The watertight meshes are quite big (~500GB) as they contain a lot of vertices and we haven't simplified them to not loose accuracy. If you need the watertight meshes, it's probably easiest to build the dataset yourself using the scripts provided in the repo.

You can also have a look at https://github.com/hjwdzh/Manifold which does a similar thing.

LMescheder commented 5 years ago

can you share the data you downloaded? I always failed to download the link provided by the author. Thank you in advance!

If you still have trouble downloading the data, please open a separate issue. Thanks!

msingh27 commented 2 years ago

Hi @LMescheder, Thanks for sharing the code repo. I used the processing step given in this repo to obtain the watertight meshes for the shapenet meshes.

However, I noticed that when I sampled points out of the mesh (using trimesh.sample) and checked its occupancy, not all points got the occupancy label "occupied" (only 60% of points got the label "occupied").

Did you also observe something similar and is this happening because the function check_mesh_contains in scripts/sample_mesh.py or trimesh.sample is approximate?

Should I use the https://github.com/hjwdzh/Manifold to obtain the watertight mesh instead (as I think all the points sampled from the mesh should have "occupied" label)?

Thanks