noether.data.datasets.cfd.shapenet_car.preprocessing¶

Preprocessing script for ShapeNet car CFD dataset.

This script processes raw VTK files containing CFD simulation data and extracts surface and volume data for machine learning applications.

Attributes¶

`logger`
`EXPECTED_SIMULATION_COUNT`
`NUM_PARAM_FOLDERS`
`EPSILON`
`PRESSURE_SURFACE_FILE`
`VELOCITY_VOLUME_FILE`
`PRESSURE_ARRAY_FILE`
`EXPECTED_CELL_TYPE`
`EXCLUDED_SIMULATIONS`
`parser`

Functions¶

`load_unstructured_grid_data`(file_path)	Load unstructured grid data from a VTK file.
`unstructured_grid_data_to_poly_data`(unstructured_grid_data)	Convert unstructured grid data to poly data by extracting the surface.
`get_sdf`(target_points, boundary_points)	Calculate signed distance field and normal directions from target points to boundary.
`get_normal`(unstructured_grid_data)	Compute normalized surface normals from unstructured grid data.
`get_simulation_relative_paths`(root)	Get relative paths to all valid simulation directories in the dataset.
`load_simulation_data`(root, simulation_relative_path)	Load and process data for a single simulation.
`save_preprocessed_data`(save_path, surface_pressure, ...)	Save preprocessed simulation data to disk.
`process_single_simulation`(root, ...)	Process a single simulation: load data, validate, and save.
`main`(root, output_dir[, continue_on_error, dry_run, ...])	Preprocess ShapeNet car dataset.

Module Contents¶

noether.data.datasets.cfd.shapenet_car.preprocessing.logger¶

noether.data.datasets.cfd.shapenet_car.preprocessing.EXPECTED_SIMULATION_COUNT = 889¶

noether.data.datasets.cfd.shapenet_car.preprocessing.NUM_PARAM_FOLDERS = 9¶

noether.data.datasets.cfd.shapenet_car.preprocessing.EPSILON = 1e-08¶

noether.data.datasets.cfd.shapenet_car.preprocessing.PRESSURE_SURFACE_FILE = 'quadpress_smpl.vtk'¶

noether.data.datasets.cfd.shapenet_car.preprocessing.VELOCITY_VOLUME_FILE = 'hexvelo_smpl.vtk'¶

noether.data.datasets.cfd.shapenet_car.preprocessing.PRESSURE_ARRAY_FILE = 'press.npy'¶

noether.data.datasets.cfd.shapenet_car.preprocessing.EXPECTED_CELL_TYPE = 'quad'¶

noether.data.datasets.cfd.shapenet_car.preprocessing.EXCLUDED_SIMULATIONS¶

noether.data.datasets.cfd.shapenet_car.preprocessing.load_unstructured_grid_data(file_path)¶

Load unstructured grid data from a VTK file.

Parameters:: file_path (pathlib.Path) – Path to the VTK file containing unstructured grid data
Returns:: vtkUnstructuredGrid object containing the loaded data
Return type:: vtk.vtkUnstructuredGrid

noether.data.datasets.cfd.shapenet_car.preprocessing.unstructured_grid_data_to_poly_data(unstructured_grid_data)¶

Convert unstructured grid data to poly data by extracting the surface.

Parameters:: unstructured_grid_data (vtk.vtkUnstructuredGrid) – vtkUnstructuredGrid object to convert
Returns:: vtkPolyData object representing the surface
Return type:: vtk.vtkPolyData

Note

The surface_filter is kept alive to maintain VTK object references.

noether.data.datasets.cfd.shapenet_car.preprocessing.get_sdf(target_points, boundary_points)¶

Calculate signed distance field and normal directions from target points to boundary.

Parameters:

target_points (numpy.typing.NDArray[numpy.float64]) – (N, 3) array of points where SDF is computed
boundary_points (numpy.typing.NDArray[numpy.float64]) – (M, 3) array of boundary surface points

Returns:

distances: (N,) array of distances to nearest boundary point
directions: (N, 3) array of normalized direction vectors to nearest boundary point

Return type:

Tuple of (distances, directions) where

noether.data.datasets.cfd.shapenet_car.preprocessing.get_normal(unstructured_grid_data)¶

Compute normalized surface normals from unstructured grid data.

Parameters:: unstructured_grid_data (vtk.vtkUnstructuredGrid) – vtkUnstructuredGrid object containing the mesh
Returns:: (N, 3) array of normalized normal vectors at each point
Raises:: RuntimeError – If NaN values are detected in the computed normals
Return type:: numpy.typing.NDArray[numpy.float64]

Note

This function modifies the input unstructured_grid_data by setting cell normals.

noether.data.datasets.cfd.shapenet_car.preprocessing.get_simulation_relative_paths(root)¶

Get relative paths to all valid simulation directories in the dataset.

The dataset is organized into 9 parameter folders (param0-param8), each containing multiple simulation subdirectories. Some simulations are excluded due to missing required files.

Parameters:

root (pathlib.Path) – Path to the root directory containing param folders

Returns:

List of Path objects representing relative paths to valid simulations

Raises:

ValueError – If the expected number of simulations is not found
FileNotFoundError – If a parameter folder does not exist

Return type:

list[pathlib.Path]

noether.data.datasets.cfd.shapenet_car.preprocessing.load_simulation_data(root, simulation_relative_path)¶

Load and process data for a single simulation.

Parameters:

root (pathlib.Path) – Root directory containing raw data
simulation_relative_path (pathlib.Path) – Relative path to the simulation directory

Returns:

Tuple of (surface_pressure, surface_position, surface_normals, mask,: exterior_points, exterior_velocity, exterior_sdf, exterior_normals)

Raises:

FileNotFoundError – If required files are missing
ValueError – If data validation fails

Return type:

tuple[numpy.typing.NDArray, numpy.typing.NDArray, numpy.typing.NDArray, numpy.typing.NDArray, numpy.typing.NDArray, numpy.typing.NDArray, numpy.typing.NDArray, numpy.typing.NDArray]

noether.data.datasets.cfd.shapenet_car.preprocessing.save_preprocessed_data(save_path, surface_pressure, surface_position, surface_normals, mask, exterior_points, exterior_velocity, exterior_sdf, exterior_normals)¶

Save preprocessed simulation data to disk.

Parameters:

save_path (pathlib.Path) – Directory where preprocessed data will be saved
surface_pressure (numpy.typing.NDArray) – Surface pressure values
surface_position (numpy.typing.NDArray) – Surface point positions
surface_normals (numpy.typing.NDArray) – Surface normal vectors
mask (numpy.typing.NDArray) – Boolean mask for valid surface points
exterior_points (numpy.typing.NDArray) – Exterior (volume) point positions
exterior_velocity (numpy.typing.NDArray) – Velocity at exterior points
exterior_sdf (numpy.typing.NDArray) – Signed distance field at exterior points
exterior_normals (numpy.typing.NDArray) – Normal vectors at exterior points

Return type:

None

noether.data.datasets.cfd.shapenet_car.preprocessing.process_single_simulation(root, simulation_relative_path, output_dir)¶

Process a single simulation: load data, validate, and save.

Parameters:

root (pathlib.Path) – Root directory containing raw data
simulation_relative_path (pathlib.Path) – Relative path to the simulation
output_dir (pathlib.Path) – Output directory for preprocessed data

Raises:

Various exceptions from data loading and validation –

Return type:

None

noether.data.datasets.cfd.shapenet_car.preprocessing.main(root, output_dir, continue_on_error=False, dry_run=False, overwrite=False)¶

Preprocess ShapeNet car dataset.

Parameters:

root (pathlib.Path) – Path to the root directory containing the raw data
output_dir (pathlib.Path) – Path to the output directory where preprocessed data will be saved
continue_on_error (bool) – If True, continue processing on errors instead of stopping
dry_run (bool) – If True, only validate data without saving
overwrite (bool) – If True, allow overwriting existing output directory

Returns:

{: ‘total’: total simulations, ‘success’: successfully processed, ‘failed’: failed to process

}

Return type:

Dictionary with statistics

Raises:

FileNotFoundError – If root directory does not exist
FileExistsError – If output directory exists and overwrite is False

noether.data.datasets.cfd.shapenet_car.preprocessing.parser¶