class test_load_ply:
from cp.tools.io import load_ply
v, f = load_ply(
os.path.join(
os.path.split(os.path.abspath(__file__))[0], "data",
"Armadillo_ascii_converted_using_meshlab.ply"))
npt.assert_equal(f.max(), len(v) - 1)
npt.assert_equal(np.setdiff1d(f.ravel(), np.arange(len(v))), np.array([]))
v2, f2 = load_ply(
os.path.join(
os.path.split(os.path.abspath(__file__))[0], "data",
"Armadillo.ply"))
# big absolute tolerance because meshlab lost many digits when
# converting to ascii
npt.assert_allclose(v2, v, atol=1e-4)
npt.assert_allclose(f2, f)
def __init__(self, ff=None):
'''
Constructor
'''
if ff is None:
ff = open('Armadillo.ply')
v, f = load_ply(ff)
ma = v.max()
mi = v.min()
med = (ma + mi) / 2
v -= med
v /= (ma - mi) / (2 * 1.7)
self.v = v
self.f = f
self.meshcp = MeshCP(v, f)
def __init__(self,ff = None):
'''
Constructor
'''
if ff is None:
ff = open('Armadillo.ply')
v,f = load_ply(ff)
ma = v.max()
mi = v.min()
med = (ma+mi)/2
v -= med
v /= (ma-mi)/(2*1.7)
self.v = v
self.f = f
self.meshcp = MeshCP(v,f)
from cp.surfaces import Mesh
# Since our mesh is a sphere, we'll take advantage of its
# parametric_plot method
from cp.surfaces import Sphere
from cp.tools.io import load_ply
from cp.build_matrices import build_interp_matrix, build_diff_matrix
# TODO: move coordinate_transform out of cp.surfaces (maybe to
# cp.tools?)
from cp.surfaces.coordinate_transform import cart2sph
PLOT = True
# Load vertices and faces, and instatiate surface
v, f = load_ply('cp/tests/data/sphere_refined.ply')
m = Mesh(v, f)
p = 3
diff_stencil_arm = 1
dim = 3
index, distance, grid, dx = m.grid(num_blocks_per_dim=41,
levels=1,
p=p,
diff_stencil_arm=diff_stencil_arm)
cp, dist, _, _ = m.closest_point(grid, index)
# The points in `grid` can be thought to be a subset of a virtual grid
# (for instance, the result of meshgrid). `ll` is the lower left
# corner of such virtual grid, and `ur` the upper right corner. The
Output
------
cp : array of shape (npoints, 4)
Coordinates of the closest points, and squareddistance from
its original point.
cp[:, 0] is the squared distance (redundant information)
cp[:, 1:] are the coordinates of the closest points
"""
N = len(index_cp)
cp = np.empty((N, 4), dtype=np.float64)
for i, (index_of_cp, point) in enumerate(izip(index_cp, grid_fine)):
face_indices_that_share_that_vertex = vertex2faces[index_of_cp]
# Much faster using np.take than fancy indexing
possible_faces = np.take(faces,
face_indices_that_share_that_vertex,
axis=0)
# Cythonize this and it gets way faster
cp[i] = FindClosestPointToTriSet(point[0], point[1], point[2],
possible_faces, vertices)
return cp
if __name__ == '__main__':
v, f = load_ply('cp/tests/data/Armadillo.ply')
m = Mesh(v, f)
v2, f2 = load_ply('cp/tests/data/eight_refined.ply')
m2 = Mesh(v2, f2)
index, grid, distance, dx = m2.grid(num_blocks=10, levels=4)
cp, dist, _, _ = m2.closest_point(index, grid)
Given by build_vertices2faces
vertices, faces : arrays, given by load_ply
Output
------
cp : array of shape (npoints, 4)
Coordinates of the closest points, and squareddistance from
its original point.
cp[:, 0] is the squared distance (redundant information)
cp[:, 1:] are the coordinates of the closest points
"""
N = len(index_cp)
cp = np.empty((N, 4), dtype=np.float64)
for i, (index_of_cp, point) in enumerate(izip(index_cp, grid_fine)):
face_indices_that_share_that_vertex = vertex2faces[index_of_cp]
# Much faster using np.take than fancy indexing
possible_faces = np.take(faces, face_indices_that_share_that_vertex, axis=0)
# Cythonize this and it gets way faster
cp[i] = FindClosestPointToTriSet(point[0], point[1], point[2], possible_faces, vertices)
return cp
if __name__ == "__main__":
v, f = load_ply("cp/tests/data/Armadillo.ply")
m = Mesh(v, f)
v2, f2 = load_ply("cp/tests/data/eight_refined.ply")
m2 = Mesh(v2, f2)
index, grid, distance, dx = m2.grid(num_blocks=10, levels=4)
cp, dist, _, _ = m2.closest_point(index, grid)