Skip to content

Commit

Permalink
Significant optimization. Added UV island generation.
Browse files Browse the repository at this point in the history
  • Loading branch information
@amb
amb committed Oct 24, 2018
1 parent 9646372 commit fd99396
Show file tree
Hide file tree
Showing 3 changed files with 92 additions and 46 deletions.
15 changes: 10 additions & 5 deletions src/mesh_segmentation/__init__.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -24,21 +24,26 @@ class MeshSegmentation(bpy.types.Operator):
items = [('assignMaterials',
"Assign materials",
"Assigns a different material for "
"each found segment")],
"each found segment"),
('assignUVs',
"Assign UV islands",
"Assigns a different UV island for "
"each found segment")
],
description = "What to do with the "
"segmentation",
default = 'assignMaterials')
default = 'assignUVs')
k = bpy.props.IntProperty(name = "Clusters",
description = "Amount of clusters",
min = 2,
default = 2)
default = 20)
delta = bpy.props.FloatProperty(name = "Delta",
description = "Set close to zero for more "
"importance on the angular "
"distance, set close to one "
"for more importance on the "
"geodesic distance.",
default = 0.03,
default = 0.5,
min = 0,
max = 1,
subtype = 'FACTOR')
Expand All @@ -48,7 +53,7 @@ class MeshSegmentation(bpy.types.Operator):
"set close to one to treat "
"concave and convex angles "
"equally.",
default = 0.15,
default = 0.5,
min = 1e-10,
max = 1,
subtype = 'FACTOR')
Expand Down
30 changes: 30 additions & 0 deletions src/mesh_segmentation/actions.py
View file
Original file line number Diff line number Diff line change
@@ -1,5 +1,6 @@
import random
import bpy
import bmesh

def assignMaterials(mesh, k, idx):
"""Assigns a random colored material for each found segment"""
Expand All @@ -17,3 +18,32 @@ def assignMaterials(mesh, k, idx):

for i, id in enumerate(idx):
mesh.polygons[i].material_index = id

def assignUVs(mesh, k, idx):
"""Assigns a UV island for each found segment"""
bpy.ops.object.mode_set(mode='EDIT')

bm = bmesh.from_edit_mesh(mesh)

# currently blender needs both layers.
uv_layer = bm.loops.layers.uv.verify()
bm.faces.layers.tex.verify()

bm.faces.ensure_lookup_table()

# create UVs from clusters
for i, id in enumerate(idx):
f = bm.faces[i]
for l in f.loops:
luv = l[uv_layer]
luv.uv.x = l.vert.co.x + id * 1.0
luv.uv.y = l.vert.co.y

bmesh.update_edit_mesh(mesh)

bpy.ops.uv.seams_from_islands()
bpy.ops.mesh.select_all(action='SELECT')
bpy.ops.uv.unwrap(method='ANGLE_BASED', margin=0.001)



93 changes: 52 additions & 41 deletions src/mesh_segmentation/segmentation.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,9 @@
import scipy.cluster
import scipy.sparse
import scipy.sparse.csgraph
import scipy.sparse.linalg

import cProfile, pstats, io

# Controls weight of geodesic to angular distance. Values closer to 0 give
# the angular distance more importance, values closer to 1 give the geodesic
Expand All @@ -16,29 +19,6 @@
# angles, values close to 1 treat convex and concave angles more equally
eta = None


class ProgressBar:

def __init__(self, steps):
self._active = (hasattr(bpy.context.window_manager,'progress_begin') and
hasattr(bpy.context.window_manager,'progress_update')and
hasattr(bpy.context.window_manager,'progress_end'))
if self._active:
self._steps = 0
self._max_steps = steps
bpy.context.window_manager.progress_begin(0, 100)
bpy.context.window_manager.progress_update(0)

def step(self):
if self._active:
self._steps += 1
bpy.context.window_manager.progress_update(self._steps /
self._max_steps)
if self._steps == self._max_steps:
bpy.context.window_manager.progress_end()
self._active = False


def _face_center(mesh, face):
"""Computes the coordinates of the center of the given face"""
center = mathutils.Vector()
Expand Down Expand Up @@ -76,7 +56,8 @@ def _create_distance_matrices(mesh, save_dists):
faces = mesh.polygons
l = len(faces)

if not save_dists and ("geo_dist_mat" in mesh and "ang_dist_mat" in mesh and
# remove False to have this actually be sometimes true
if False and not save_dists and ("geo_dist_mat" in mesh and "ang_dist_mat" in mesh and
"geo_dist_avg" in mesh and "ang_dist_sum" in mesh and
"num_adj" in mesh and "use_eta_list" in mesh):
# the matrices are already calculated, we only have to load and
Expand All @@ -97,14 +78,13 @@ def _create_distance_matrices(mesh, save_dists):
# map from edge-key to adjacent faces
adj_faces_map = {}
# find adjacent faces by iterating edges
progress = ProgressBar(steps = l)
print("find adjacent faces")
for index, face in enumerate(faces):
for edge in face.edge_keys:
if edge in adj_faces_map:
adj_faces_map[edge].append(index)
else:
adj_faces_map[edge] = [index]
progress.step()

# average G and cumulated A
avgG = 0
Expand All @@ -117,7 +97,7 @@ def _create_distance_matrices(mesh, save_dists):
Gcol = []
Gval = []
# iterate adjacent faces and calculate distances
progress = ProgressBar(steps = len(adj_faces_map))
print("iterate adjacent faces and calculate distances")
for edge, adj_faces in adj_faces_map.items():
if len(adj_faces) == 2:
i = adj_faces[0]
Expand Down Expand Up @@ -151,9 +131,8 @@ def _create_distance_matrices(mesh, save_dists):
elif len(adj_faces) > 2:
print("Edge with more than 2 adjacent faces: " + str(adj_faces) + "!")

progress.step()

# create sparse matrices
print("create sparse matrices")
# matrix of geodesic distances
G = scipy.sparse.csr_matrix((Gval, (Grow, Gcol)), shape=(l, l))
# matrix of angular distances
Expand All @@ -177,8 +156,7 @@ def _create_affinity_matrix(mesh):

l = len(mesh.polygons)
print("mesh_segmentation: Creating distance matrices...")
G, A, avgG, sumA, num_adj, use_eta_list = _create_distance_matrices(mesh,
False)
G, A, avgG, sumA, num_adj, use_eta_list = _create_distance_matrices(mesh, False)

# scale needed angular distances with eta
for indices in use_eta_list:
Expand All @@ -193,7 +171,9 @@ def _create_affinity_matrix(mesh):

print("mesh_segmentation: Finding shortest paths between all faces...")
# for each non adjacent pair of faces find shortest path of adjacent faces
#W = scipy.sparse.csgraph.dijkstra(G + A, unweighted = True, directed = False)
W = scipy.sparse.csgraph.dijkstra(G + A, directed = False)
print("indices")
inf_indices = numpy.where(numpy.isinf(W))
W[inf_indices] = 0

Expand All @@ -210,14 +190,11 @@ def _create_affinity_matrix(mesh):

def _initial_guess(Q, k):
"""Computes an initial guess for the cluster-centers"""
n = Q.shape[0]
min_value = 2
min_indices=(-1,-1)
for (i,j), value in numpy.ndenumerate(Q):
if i != j and value < min_value:
min_value = Q[i,j]
min_indices = (i,j)
min_value = numpy.min(Q)
loc = numpy.argmin(Q)
min_indices = (loc // Q.shape[1], loc % Q.shape[0])

n = Q.shape[0]
chosen = [min_indices[0], min_indices[1]]
for _ in range(2,k):
min_max = float("inf")
Expand All @@ -239,38 +216,72 @@ def segment_mesh(mesh, k, coefficients, action):
action for each cluster
"""

# profiling
pr = cProfile.Profile()
pr.enable()

# set coefficients
global delta
global eta
delta, eta = coefficients

# affinity matrix
W = _create_affinity_matrix(mesh)
print("mesh_segmentation: Calculating graph laplacian...")
print("mesh_segmentation: Calculating graph laplacian with shape:", W.shape)

# degree matrix
Dsqrt = numpy.diag([math.sqrt(1/entry) for entry in W.sum(1)])

# graph laplacian
L = Dsqrt.dot(W.dot(Dsqrt))
print("dot products")
Dsqrt_s = scipy.sparse.csr_matrix(Dsqrt)
W_s = scipy.sparse.csr_matrix(W)
L = Dsqrt_s.dot(W_s.dot(Dsqrt_s))

print("mesh_segmentation: Calculating eigenvectors...")

# get eigenvectors
l,V = scipy.linalg.eigh(L, eigvals = (L.shape[0] - k, L.shape[0] - 1))
l,V = scipy.sparse.linalg.eigsh(L)
print(l.shape, V.shape, L.shape)

# normalize each column to unit length
V = V / [numpy.linalg.norm(column) for column in V.transpose()]

print("mesh_segmentation: Preparing kmeans...")

# compute association matrix
Q = V.dot(V.transpose())

# compute initial guess for clustering
print("mesh_segmentation: Initial guess...")
initial_clusters = _initial_guess(Q, k)

print("mesh_segmentation: Applying kmeans...")

# apply kmeans
cluster_res,_ = scipy.cluster.vq.kmeans(V, V[initial_clusters,:])
print(cluster_res.shape, V.shape)

# 1. k-means clustering on curvature (from addon)
# 2. make seams
# 3. unwrap
# 4. mark all islands with loops (more than 1 borders)
# 5. connect all border chains into the closest other chain (shortest path mark seam)
# 6. repeat 3

# get identification vector
idx,_ = scipy.cluster.vq.vq(V, cluster_res)

print("mesh_segmentation: Done clustering!")

# end profile, print results
pr.disable()
s = io.StringIO()
sortby = 'cumulative'
ps = pstats.Stats(pr, stream=s)
ps.strip_dirs().sort_stats(sortby).print_stats()
print(s.getvalue())

# perform action with the clustering result
if action:
action(mesh, k, idx)

0 comments on commit fd99396

Please sign in to comment.