This project contains PyTorch implementation of Cohen, Taco, et al. "Gauge Equivariant Convolutional Networks and the Icosahedral CNN." with an example. Optionally the project can be run using docker with the provided dockerfile

Usage

The project has been implemented on python and can be run using Dockerfile

1. Build Docker Deploy Image

docker build --network=host \
  --build-arg USER_ID=$(id -u) \
  --build-arg GROUP_ID=$(id -g) \
  --build-arg USER=docker_$USER \
  -t githubicosahedralcnndocker:v1 . 

2. Run Docker Image

docker run --network host \
  --gpus all -it --rm --user $(id -u):$(id -g) \
  --shm-size=8G \
  -e HOST_HOSTNAME=`hostname` \
  --mount type=bind,src=<>,dst=<> \
  --workdir <> githubicosahedralcnndocker:v1 bash

3. Training Example

python3 examples/mnist/run.py 

Projects

If you like this project you also might be interested in other projects which use IcosahedralCNN as remeshing stage for mesh generation.