def pyoctree_volume(mesh):
try:
from pyoctree import pyoctree
except ImportError:
pytest.skip("pyoctree not available")
else:
vertices = np.asarray(mesh.points, dtype=np.float64, order="C")
faces = np.asarray(mesh.cells_dict["triangle"],
dtype=np.int32,
order="C")
octree = pyoctree.PyOctree(vertices, faces)
return PyOctreeWrapper(vertices, faces, octree)
def _in_volume_ray(points, volume):
""" Uses pyoctree's raycsasting to test if points are within a given
CATMAID volume.
"""
tree = getattr(volume, 'pyoctree', None)
if not tree:
# Create octree from scratch
tree = pyoctree.PyOctree(
np.array(volume.vertices, dtype=float, order='C'),
np.array(volume.faces, dtype=np.int32, order='C'))
volume.pyoctree = tree
# Get min max of volume
mx = np.array(volume.vertices).max(axis=0)
mn = np.array(volume.vertices).min(axis=0)
# Get points outside of bounding box
out = (points > mx).any(axis=1) | (points < mn).any(axis=1)
isin = ~out
in_points = points[~out]
# Perform ray intersection on points inside bounding box
rayPointList = np.array([[[p[0], p[1], mn[2]], [p[0], p[1], mx[2]]]
for p in in_points],
dtype=np.float32)
# Unfortunately rays are bidirectional -> we have to filter intersections
# to those that occur "above" the point we are querying
intersections = [
len([
i for i in tree.rayIntersection(ray) if i.p[2] >= in_points[k][2]
]) for k, ray in enumerate(rayPointList)
]
# Count intersections and return True for odd counts
# [i % 2 != 0 for i in intersections]
isin[~out] = np.remainder(list(intersections), 2) != 0
return isin
def in_volume_pyoc(points: np.ndarray,
volume: Volume,
n_rays: Optional[int] = 1) -> Sequence[bool]:
"""Use pyoctree's raycasting to test if points are within a given volume."""
if isinstance(n_rays, type(None)):
n_rays = 1
if not isinstance(n_rays, (int, np.integer)):
raise TypeError(f'n_rays must be integer, got "{type(n_rays)}"')
if n_rays <= 0:
raise ValueError('n_rays must be > 0')
tree = getattr(volume, 'pyoctree', None)
if not tree:
# Create octree from scratch
tree = pyoctree.PyOctree(
np.array(volume.vertices, dtype=float, order='C'),
np.array(volume.faces, dtype=np.int32, order='C'))
volume.pyoctree = tree # type: ignore
# Get min max of volume
mx = np.array(volume.vertices).max(axis=0)
mn = np.array(volume.vertices).min(axis=0)
dm = mx - mn
# Remove points outside of bounding box
is_out = (points > mx).any(axis=1) | (points < mn).any(axis=1)
# Cast rays
# There is no point of vectorizing this because pyoctree's rayIntersection
# does only take a single ray at a time...
for i in range(n_rays):
# Process only point that we think could be in
in_points = points[~is_out]
# If no in points left, break out
if in_points.size == 0:
break
# Pick a random point inside the volumes bounding box as origin
origin = np.random.rand(3) * dm + mn
# Generate ray point list:
rayPointList = np.array([[origin, p] for p in in_points],
dtype=np.float32)
# Get intersections and extract coordinates of intersection
intersections = [
np.array([i.p for i in tree.rayIntersection(ray)])
for ray in rayPointList
]
# In a few odd cases we can get the multiple intersections at the exact
# same coordinate (something funny with the faces).
unique_int = [
np.unique(np.round(i), axis=0) if np.any(i) else i
for i in intersections
]
# Unfortunately rays are bidirectional -> we have to filter intersections
# to those that occur "above" the point we are querying
unilat_int = [
i[i[:, 2] >= p] if np.any(i) else i
for i, p in zip(unique_int, in_points[:, 2])
]
# Count intersections
int_count = [i.shape[0] for i in unilat_int]
# Get even (= outside volume) numbers of intersections
is_even = np.remainder(int_count, 2) == 0
# Set outside points
is_out[~is_out] = is_even
return ~is_out
def texture_vis(model, image, data, v_t, mask):
mesh = copy(_TEMPLATE_MESH)
my_tex = np.zeros((1000, 1000, image.shape[2]), dtype=image.dtype)
#cmesh = Mesh(_os.path.join(_os.path.dirname(__file__),
# 'template-bodyparts-corrected-labeled-split5.ply'))
#mesh.vc = cmesh.vc.copy()
mesh.vc = [0.91, 0.67, 0.91]
model.betas[:len(data['betas'])] = data['betas']
model.pose[:] = data['pose']
model.trans[:] = data['trans']
mesh.v = v_t
#pdb.set_trace()
mesh.v = np.array(
_verts_core(model.pose, mesh.v, model.J, model.weights,
model.kintree_table))
#pdb.set_trace()
#new_v_t = _verts_core(model.pose, v_t, model.J, model.weights, model.kintree_table)
#mesh.v = mesh.v + np.array(new_v_t)
largura, altura = (image.shape[1], image.shape[0])
data['cam_c'] = np.array([largura, altura]) / 2.
tex_vt, tex_faces = Template_tex()
vert = mesh.v + np.array(data['trans']).reshape((1, 3))
faces = np.array(mesh.f.copy(), dtype='int32')
tree = _pyoctree.PyOctree(vert, faces)
cp = np.zeros((vert.shape[0], 4))
#my_map_partes = np.zeros(vert.shape[0]).astype(int)
for i, f in enumerate(mesh.f):
#if ((i < 300) or (i > 5046)):
# continue
I_VIS = True
face_media = (vert[f[0]] + vert[f[1]] + vert[f[2]]) / 3.0
rayList = np.array([np.zeros(3), face_media], dtype=np.float32)
intersectionFound = tree.rayIntersection(rayList)
if intersectionFound:
if dist(intersectionFound[0].p, face_media) > 0.0001:
I_VIS = False
if I_VIS:
tri1 = np.zeros((3, 2), np.float32)
tri2 = np.zeros((3, 2), np.float32)
tri1[0, 0] = (vert[f[0]][1] *
data['f']) / (vert[f[0]][2]) + data['cam_c'][1]
tri1[0, 1] = (vert[f[0]][0] *
data['f']) / (vert[f[0]][2]) + data['cam_c'][0]
tri1[1, 0] = (vert[f[1]][1] *
data['f']) / (vert[f[1]][2]) + data['cam_c'][1]
tri1[1, 1] = (vert[f[1]][0] *
data['f']) / (vert[f[1]][2]) + data['cam_c'][0]
tri1[2, 0] = (vert[f[2]][1] *
data['f']) / (vert[f[2]][2]) + data['cam_c'][1]
tri1[2, 1] = (vert[f[2]][0] *
data['f']) / (vert[f[2]][2]) + data['cam_c'][0]
cp[f[0]][0] = 1.0
cp[f[0]][1] = tri1[0, 0]
cp[f[0]][2] = tri1[0, 1]
cp[f[0]][3] = vert[f[0]][2]
cp[f[1]][0] = 1.0
cp[f[1]][1] = tri1[1, 0]
cp[f[1]][2] = tri1[1, 1]
cp[f[1]][3] = vert[f[1]][2]
cp[f[2]][0] = 1.0
cp[f[2]][1] = tri1[2, 0]
cp[f[2]][2] = tri1[2, 1]
cp[f[2]][3] = vert[f[2]][2]
tri2[0, 1] = (tex_vt[tex_faces[i][0]][0]) * 1000
tri2[0, 0] = 1000 - (tex_vt[tex_faces[i][0]][1]) * 1000
tri2[1, 1] = (tex_vt[tex_faces[i][1]][0]) * 1000
tri2[1, 0] = 1000 - (tex_vt[tex_faces[i][1]][1]) * 1000
tri2[2, 1] = (tex_vt[tex_faces[i][2]][0]) * 1000
tri2[2, 0] = 1000 - (tex_vt[tex_faces[i][2]][1]) * 1000
'''my_map_partes[f[0]] = (image[int(tri2[0,0]),int(tri2[0,1]),0])/15 if (image[int(tri2[0,0]),int(tri2[0,1]),0])/15 > 0 else my_map_partes[f[0]]
my_map_partes[f[1]] = (image[int(tri2[1,0]),int(tri2[1,1]),0])/15 if (image[int(tri2[1,0]),int(tri2[1,1]),0])/15 > 0 else my_map_partes[f[1]]
my_map_partes[f[2]] = (image[int(tri2[2,0]),int(tri2[2,1]),0])/15 if (image[int(tri2[2,0]),int(tri2[2,1]),0])/15 > 0 else my_map_partes[f[2]] '''
#if image[int(tri2[0,0]),int(tri2[0,1]),0] == 15:
#warp_tri(tri2,tri2,image,my_tex)
#tex_vt[tex_faces[i][0] - 1][1]
#print tri1
warp_tri(tri1, tri2, image, my_tex)
#cv2.imshow('ImageWindow', my_tex)
#cv2.waitKey()
#np.savetxt('test.txt', my_map_partes , delimiter='\n')
return my_tex, cp
def _in_volume_ray(points, volume, multi_ray=False):
"""Use pyoctree's raycsasting to test if points are within volume."""
tree = getattr(volume, 'pyoctree', None)
if not tree:
# Create octree from scratch
tree = pyoctree.PyOctree(
np.array(volume.vertices, dtype=float, order='C'),
np.array(volume.faces, dtype=np.int32, order='C'))
volume.pyoctree = tree
# Remove points outside of bounding box
isin = _in_bbox(points, volume)
in_points = points[isin]
# mx = np.array(volume.vertices).max(axis=0)
mn = np.array(volume.vertices).min(axis=0)
# Perform ray intersection on points inside bounding box
rayPointList = np.array([[[p[0], mn[1], mn[2]], p] for p in in_points],
dtype=np.float32)
# Get intersections and extract coordinates of intersection
intersections = [
np.array([i.p for i in tree.rayIntersection(ray)])
for ray in rayPointList
]
# In a few odd cases we can get the multiple intersections at the exact
# same coordinate (something funny with the faces).
unique_int = [
np.unique(np.round(i), axis=0) if np.any(i) else i
for i in intersections
]
# Unfortunately rays are bidirectional -> we have to filter intersections
# to those that occur "above" the point we are querying
unilat_int = [
i[i[:, 2] >= p] if np.any(i) else i
for i, p in zip(unique_int, in_points[:, 2])
]
# Count intersections
int_count = [i.shape[0] for i in unilat_int]
# Get odd (= in volume) numbers of intersections
is_odd = np.remainder(int_count, 2) != 0
# If we want to play it safe, run the above again with two additional rays
# and find a majority decision.
if multi_ray:
# Run ray from left back
rayPointList = np.array([[[mn[0], p[1], mn[2]], p] for p in in_points],
dtype=np.float32)
intersections = [
np.array([i.p for i in tree.rayIntersection(ray)])
for ray in rayPointList
]
unique_int = [
np.unique(i, axis=0) if np.any(i) else i for i in intersections
]
unilat_int = [
i[i[:, 0] >= p] if np.any(i) else i
for i, p in zip(unique_int, in_points[:, 0])
]
int_count = [i.shape[0] for i in unilat_int]
is_odd2 = np.remainder(int_count, 2) != 0
# Run ray from lower left
rayPointList = np.array([[[mn[0], mn[1], p[2]], p] for p in in_points],
dtype=np.float32)
intersections = [
np.array([i.p for i in tree.rayIntersection(ray)])
for ray in rayPointList
]
unique_int = [
np.unique(i, axis=0) if np.any(i) else i for i in intersections
]
unilat_int = [
i[i[:, 1] >= p] if np.any(i) else i
for i, p in zip(unique_int, in_points[:, 1])
]
int_count = [i.shape[0] for i in unilat_int]
is_odd3 = np.remainder(int_count, 2) != 0
# Find majority consensus
is_odd = is_odd.astype(int) + is_odd2.astype(int) + is_odd3.astype(int)
is_odd = is_odd >= 2
isin[isin] = is_odd
return isin
#Point coordinates is an Nx3 array representing the coordinates of the points
#connectivity is an Nx3 integer array describing which points are connected
mesh = trimesh.load_mesh('transectShape.ply')
f = np.array(mesh.faces).astype(np.int32)
v = np.array(mesh.vertices)
minX = np.min(v[:, 0])
minY = np.min(v[:, 1])
v[:, 0] += -minX
v[:, 1] += -minY
pointCoords = v
connectivity = f
tree = ot.PyOctree(pointCoords, connectivity)
# Choose some starting points, for example the positions of the cameras
import pandas as pd
pose = pd.read_csv(
'C:/Users/haavasl/PycharmProjects/Testing/venv/Code/nav_file_for_tautra_new.txt'
)
X = pose['X'].values.reshape((-1, 1))
Y = pose['Y'].values.reshape((-1, 1))
Z = pose['Z'].values.reshape((-1, 1))
xyz_start = np.zeros((X.shape[0], 3), dtype=np.float32)
xyz_end = np.zeros((X.shape[0], 3), dtype=np.float32)