This repository contains code to generate the data used by the sibling repo for this project.
pip install -r requirements.txt
You will also need to have Docker installed to run
cuda_voxelizer. You will also
need to compile and make available
ManifoldPlus in the bin folder.
Finally, you will need to obtain access to a dataset of meshes. Existing
artifacts can be pulled using dvc pull if you have access to the existing
artifact repository S3 bucket at s3://cornell-mfd64/dvc/text2building_data_2.
Contact the owner of this repository to receive access.
If you have access, all generated artifacts can be directly downloaded instead of generated, using the following command:
dvc pull
If you are unable to obtain access to the artifact repository, the code here can
be used to generate new artifacts from a custom dataset. Simply populate
citydata/buildingnet/meshes with any number of OBJ meshes, and run the following command:
dvc repro
This will compute all artifacts according to the DAG illustrated above.
- [] Dockerize ManifoldPlus
- [] Make pipeline more generic to other datasets