-
Notifications
You must be signed in to change notification settings - Fork 6
/
Cartoon.py
133 lines (105 loc) · 5.16 KB
/
Cartoon.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
import numpy as np
import igl
import utils
import matplotlib.pyplot as plt
import os
from correspondence import Shape
from correspondence import CorrespondenceSolver
try:
ps = __import__('polyscope')
print('Visualization with polyscope.')
visualization = True
except ImportError:
print("Module polyscope not found. No visualization of result.")
visualization = False
if __name__ == "__main__":
readPath = 'data/'
writePath = 'results/'
if not os.path.exists(writePath):
os.mkdir(writePath)
# example1
# nameA = 'headA'
# nameB = 'headB'
# example2
nameA = 'homerA'
nameB = 'homerB'
bending_weight = 1e-5
kmin = 5
kmax = 100
precise = False # convert the final vertex map to a vertex-to-point map
exists_gt = True # groundtruth exists
# load the groundtruth correspondences for measuring the error
groundtruth = np.loadtxt(readPath + 'gt_' + nameA + '_' + nameB + '.txt').astype(int)
# load 5 landmark correspondences
landmarksA = np.loadtxt(readPath + nameA + '_landmarks.txt').astype(int)
landmarksB = groundtruth[landmarksA]
vA, textA, n, fA, m, b = igl.read_obj(readPath + nameA + '.obj')
vB, textB, n, fB, m, b = igl.read_obj(readPath + nameB + '.obj')
shapeA = Shape(vA, fA, name=nameA)
shapeB = Shape(vB, fB, name=nameB)
# compute correspondences using the elastic eigenmodes (elasticBasis)
Solv_elastic = CorrespondenceSolver(shapeA, shapeB, kmin=kmin, kmax=kmax, bending_weight=bending_weight,
elasticBasis=True)
# create an initial functional map (kmin x kmin) by aligning the basis functions on landmarks
C_init = Solv_elastic.CinitfromLandmarks(landmarksA, landmarksB)
# compute final correspondences by an iterative procedure
P, C = Solv_elastic.computeCorrespondence(C=C_init)
# convert mapping matrix to indices vA->vB[corr_ours]
corr_ours = P.toarray()
corr_ours = np.nonzero(corr_ours.T)[1]
np.savetxt(writePath + shapeA.name + '_' + shapeB.name + '_ElasticBasisresult.txt', corr_ours, fmt='%d')
if precise:
P_prec = Solv_elastic.preciseMap(C)
np.save(writePath + shapeA.name +'_' + shapeB.name +'ElasticPrecisemap.npy', P_prec, allow_pickle = True)
# use the eigenfunctions of LB operator as a comparison (this method corresponds to ZoomOut)
Solv_LB = CorrespondenceSolver(shapeA, shapeB, kmin=kmin, kmax=kmax, LB=True)
C_init = Solv_LB.CinitfromLandmarks(landmarksA, landmarksB)
P_LB, C_LB = Solv_LB.computeCorrespondence(C=C_init)
# #convert mapping matrix to indices vA->vB[corr_LB]
corr_LB = P_LB.toarray()
corr_LB = np.nonzero(corr_LB.T)[1]
if precise:
P_prec_LB = Solv_LB.preciseMap(C_LB)
np.save(writePath + shapeA.name +'_' + shapeB.name +'_LBPrecisemap.npy', P_prec_LB, allow_pickle = True)
np.savetxt(writePath + shapeA.name + '_' + shapeB.name + '_LBBasisresult.txt', corr_LB, fmt='%d')
print('saved computed correspondence in result folder')
if exists_gt:
# compute geodesic error of final resulta
distmatrix = utils.geodesic_distmat_dijkstra(shapeB.v, shapeB.f)
distmatrix = distmatrix / np.sqrt(np.sum(shapeB.mass))
# error of Ours
dist = distmatrix[corr_ours, groundtruth]
errOurs, percOurs = utils.compute_percentageError(dist, maxdist=0.1, step=0.01)
# error of LB approach
dist = distmatrix[corr_LB, groundtruth]
errZO, percZO = utils.compute_percentageError(dist, maxdist=0.1, step=0.01)
fig, ax = plt.subplots()
ax.get_xaxis().set_visible(True)
ax.get_yaxis().set_visible(True)
ax.set_ylabel('% correspondences', labelpad=10)
ax.set_xlabel('geodesic error', labelpad=10)
ax.set_xlim((0, 0.09))
ax.set_ylim((0, 102))
ax.plot(errOurs, percOurs * 100, c='red', label='elasticBasis', linewidth=3)
ax.plot(errZO, percZO * 100, c='blue', label='LBBasis', linewidth=3)
label_params = ax.get_legend_handles_labels()
label_params[1].reverse()
label_params[0].reverse()
plt.grid()
fig.legend()
fig.savefig(writePath + 'errorPlot' + shapeA.name + '_' + shapeB.name)
print("saved error plot in results folder")
########visualize results#######
if visualization:
ps.init()
source_mesh = ps.register_surface_mesh("source shape", shapeA.v, shapeA.f, smooth_shade=True)
target_mesh = ps.register_surface_mesh("target shape", shapeB.v, shapeB.f, smooth_shade=True)
# normal transfer
target_mesh.add_color_quantity("normals", shapeB.normals, enabled=True)
source_mesh.add_color_quantity("elastic Basis pullback normals", shapeB.normals[corr_ours], enabled=True)
source_mesh.add_color_quantity("LB Basis pullback normals", shapeB.normals[corr_LB], enabled=False)
target_mesh.set_position(np.array([0, 0, 1]))
if precise:
source_mesh.add_color_quantity("elastic precise map pullback normals", P_prec.dot(shapeB.normals), enabled=True)
source_mesh.add_color_quantity("LB precise map pullback normals", P_prec_LB.dot(shapeB.normals), enabled=False)
ps.show()