Skip to content
Dual Convolution Mesh Network
Branch: master
Clone or download

Latest commit

Fetching latest commit…
Cannot retrieve the latest commit at this time.


Type Name Latest commit message Commit time
Failed to load latest commit information.
doc smaller teaser image Feb 11, 2020 Initial Commit Feb 11, 2020

Dual Convolution Mesh Network (DCM Net)

prediction example

Coming soon...

Please stay tuned; we are currently working hard to get the code out quickly.


All of our dependencies can be installed with conda or pip.

  • Python 3.7
  • Open3D
  • PyTorch 1.1 Cuda 10.0
  • TensorboardX (Tensorflow and Tensorboard are unfortunately also needed to install this)
  • Our fork of PyTorch Geometric (with its accompanying libraries as torch_scatter, torch_cluster, torch_sparse)
  • tqdm

Since we adapted PyTorch Geometric to enable graph level support, you need to install our fork as follows:

cd pytorch_geometric
python install


Please refer to to get access to the ScanNet dataset. Our method relies on the .ply as well as the .labels.ply files.

Start a new training:

python \

Resume a training:

python \

Reproduce the scores of our paper:

python \
-c experiments/EXPERIMENT_NAME.json \
-r paper_checkpoints/EXPERIMENT_NAME.pth \
You can’t perform that action at this time.