neuralop.data.datasets.car_cfd_dataset.CarCFDDataset

class neuralop.data.datasets.car_cfd_dataset.CarCFDDataset(root_dir: str | Path, n_train: int = 1, n_test: int = 1, query_res: List[int] = [32, 32, 32], download: bool = True)[source]

CarCFDDataset is a processed version of the dataset introduced in [1], which encodes a triangular mesh over the surface of a 3D model car and provides the air pressure at each centroid and vertex of the mesh when the car is placed in a simulated wind tunnel with a recorded inlet velocity. In our case, inputs are a signed distance function evaluated over a regular 3D grid of query points, as well as the inlet velocity. Outputs are pressure values at each centroid of the triangle mesh.

CarCFDDataset inherits from MeshDataModule, which requires the optional open3d dependency. See Fast 3D spatial computing with Open3D for more information.

We also add additional manifest files to split the provided examples into training and testing sets, as well as remove instances that are corrupted.

Data is also stored on Zenodo: https://zenodo.org/records/13936501

Parameters:

root_dirUnion[str, Path]

root directory at which data is stored.

n_trainint, optional

Number of training instances to load, by default 1

n_testint, optional

Number of testing instances to load, by default 1

query_resList[int], optional

Dimension-wise resolution of signed distance function (SDF) query cube, by default [32,32,32]

downloadbool, optional

Whether to download data from Zenodo, by default True

Attributes:

train_loader: torch.utils.data.DataLoader

dataloader of training examples

test_loader: torch.utils.data.DataLoader

dataloader of testing examples

References

[1]

:

Umetani, N. and Bickel, B. (2018). “Learning three-dimensional flow for interactive

aerodynamic design”. ACM Transactions on Graphics, 2018. https://dl.acm.org/doi/10.1145/3197517.3201325.