The PolyData class: point cloud, wireframe and triangle meshes

This notebook describes the PolyData class, the common class for point clouds, wireframe and triangle meshes.

• Load and create PolyData

• Plot PolyData with PyVista

• Add a signal

• Add landmarks

• Add control points

• Save the object and load it back

PolyData creation

 There are two ways to create a PolyData object:

• from a file

• explicitly with points, edges (optional), triangles (optional)

import pyvista as pv

import skshapes as sks

cpos = [(-1.6657788922829617, 7.472045340108491, 3.9439767221656665),
 (0.8380894707515836, -0.003572508692741394, -0.002311795949935913),
 (0.9587598899863243, 0.2457800580099367, 0.14272924170625823)]


# Load a mesh from a .ply file
mesh = sks.PolyData("../data/human/human.ply")

# Extract the points, edges and triangles from the mesh
points, edges, triangles = mesh.points, mesh.edges, mesh.triangles

# Create a point cloud, a wireframe and a surface mesh from the points, edges and triangles
points_cloud = sks.PolyData(points=points)
wireframe_mesh = sks.PolyData(points=points, edges=edges)
surface_mesh = sks.PolyData(points=points, triangles=triangles)

# Create a plotter with 3 subplots and display the point cloud, wireframe and surface mesh
# PolyData objects can be converted to pyvista objects with the to_pyvista method
plotter = pv.Plotter(shape=(1, 3))
plotter.subplot(0, 0)
plotter.add_text("Points cloud", font_size=24)
plotter.add_mesh(points_cloud.to_pyvista(), color="red", render_points_as_spheres=True)
plotter.camera_position = cpos
plotter.subplot(0, 1)
plotter.add_text("Wireframe", font_size=24)
plotter.add_mesh(wireframe_mesh.to_pyvista(), color="k", line_width=2)
plotter.camera_position = cpos
plotter.subplot(0, 2)
plotter.add_text("Triangle mesh", font_size=24)
plotter.add_mesh(surface_mesh.to_pyvista(), color="tan")
plotter.camera_position = cpos
plotter.show()

Static Scene

Interactive Scene

Add signals

• Signals can be added to the PolyData object through point_data, edge_data or triangle_data.

• The size of the array in the first dimension is expected to match n_points, n_edges or`n_triangles`.

import torch

# Add some user-defined data to the mesh
surface_mesh.triangle_data["signal_triangles"] = torch.rand(surface_mesh.n_triangles, 3)
surface_mesh.edge_data["signal_edges"] = torch.rand(surface_mesh.n_edges, 3, 3)
surface_mesh.point_data["signal_points"] = surface_mesh.points[:, 0]

# Some features are already available in the mesh object
surface_mesh.triangle_data["centers"] = surface_mesh.triangle_centers
surface_mesh.triangle_data["areas"] = surface_mesh.triangle_areas
surface_mesh.edge_data["centers"] = surface_mesh.edge_centers
surface_mesh.edge_data["lengths"] = surface_mesh.edge_lengths

# Call the plot method display information about the signal data (size, type, device)
print("Triangle_data:")
print(surface_mesh.triangle_data)

# Point_data can be used in the plot method to color the surface mesh
surface_mesh.plot(scalars="signal_points", cpos=cpos)

Static Scene

Interactive Scene

Triangle_data:
DataAttributes Object with attributes:
- signal_triangles: torch.Size([250, 3]), torch.float32
- centers: torch.Size([250, 3]), torch.float32
- areas: torch.Size([250]), torch.float32
Number of elements: 250
Device: cpu

Add landmarks

• Landmarks are distinguished points of the PolyData

• They can be selected via an interactive application (using vedo)

• Landmarks can be accessed as (n_landmarks, 3) tensors or list of indices

if not pv.BUILDING_GALLERY:
    # If not in the gallery, we can use vedo to open the landmark setter
    # Setting the default backend to vtk is necessary when running in a notebook
    import vedo
    vedo.settings.default_backend= 'vtk'
    sks.LandmarkSetter(surface_mesh).start()
else:
    # Set the landmarks manually
    surface_mesh.landmark_indices = [122,  82,  89,  28,  27]

plotter = pv.Plotter()
plotter.add_mesh(surface_mesh.to_pyvista(), color="tan")
plotter.add_points(surface_mesh.landmark_points.numpy(), color="red", point_size=10, render_points_as_spheres=True)
plotter.camera_position = cpos
plotter.add_title(f"Landmarks : {surface_mesh.landmark_indices}")
plotter.show()

Static Scene

Interactive Scene

Add control points

• Control points are represented by another PolyData, attached to the first

• The bounding_grid method offers a standard way to define control points as regular grids

surface_mesh.control_points = surface_mesh.bounding_grid(N=10, offset=0.05)

plotter = pv.Plotter()
plotter.add_mesh(surface_mesh.to_pyvista(), color="tan")
plotter.add_mesh(surface_mesh.control_points.to_pyvista(), color="green", opacity=0.9)
plotter.camera_position = cpos
plotter.show()

Static Scene

Interactive Scene

Save and load

• PolyData can be saved to the disk

• If the file format is .vtk, signals and landmarks persist when loaded back

• it is also possible to save in .stl or .obj but signals/landmarks will be lost

• Control points must be saved separately

extension = ".vtk"

from pathlib import Path

filename_mesh = "tmp" + extension
filename_control_points = "tmp_control_points" + extension

surface_mesh.save(filename_mesh)
surface_mesh.control_points.save(filename_control_points)

loaded_mesh = sks.PolyData(filename_mesh)
loaded_mesh.control_points = sks.PolyData(filename_control_points)

plotter = pv.Plotter()
plotter.add_mesh(loaded_mesh.to_pyvista(), scalars="signal_points", cmap="viridis")
plotter.add_points(loaded_mesh.landmark_points.numpy(), color="red", point_size=10, render_points_as_spheres=True)
plotter.add_mesh(loaded_mesh.control_points.to_pyvista(), color="green", opacity=0.9)
plotter.camera_position = cpos
plotter.add_title(f"Landmarks : {loaded_mesh.landmark_indices}")
plotter.show()

print("Triangle_data:")
print(loaded_mesh.triangle_data)

Path.unlink(Path(filename_mesh))
Path.unlink(Path(filename_control_points))

Static Scene

Interactive Scene

Triangle_data:
DataAttributes Object with attributes:
- signal_triangles: torch.Size([250, 3]), torch.float32
- centers: torch.Size([250, 3]), torch.float32
- areas: torch.Size([250]), torch.float32
Number of elements: 250
Device: cpu