-
Notifications
You must be signed in to change notification settings - Fork 9
/
computeChamfer.py
263 lines (238 loc) · 8.78 KB
/
computeChamfer.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
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
import argparse
import glob
import re
import os
import pymeshlab
import trimesh
import glob
import numpy as np
import time
import os
import matplotlib.pyplot as plt
import matplotlib.colors as colors
import matplotlib.cm as cmx
from typing import Dict,Union
from pykdtree.kdtree import KDTree
import sys
folder = None
normalize = None
class Tee(object):
def __init__(self, name, mode):
self.file = open(name, mode)
self.stdout = sys.stdout
sys.stdout = self
def __del__(self):
sys.stdout = self.stdout
self.file.close()
def write(self, data):
self.file.write(data)
self.stdout.write(data)
def flush(self):
self.file.flush()
def load_point_cloud(path,num_points = -1):
ms = pymeshlab.MeshSet()
ms.load_new_mesh(path)
mesh = ms.current_mesh()
coords = mesh.vertex_matrix().astype('float32')
normals = mesh.vertex_normal_matrix().astype('float32')
if num_points > 0 and num_points < coords.shape[0]:
idx = np.random.permutation(coords.shape[0])[:num_points]
coords = np.ascontiguousarray(coords[idx])
normals = np.ascontiguousarray(normals[idx])
print(coords.shape)
return coords, normals
def get_youngest_file(files):
""" Find and return the oldest file of input file names.
Only one wins tie. Values based on time distance from present.
Use of `_invert` inverts logic to make this a youngest routine,
to be used more clearly via `get_youngest_file`.
"""
# Check for empty list.
if not files:
return None
if len(files) == 1:
return files[0]
# Raw epoch distance.
now = time.time()
# Select first as arbitrary sentinel file, storing name and age.
youngest = files[0], now - os.path.getctime(files[0])
# Iterate over all remaining files.
for f in files[1:]:
age = now - os.path.getctime(f)
if age < youngest[1]:
# Set new oldest.
youngest = f, age
# Return just the name of oldest file.
return youngest[0]
def get_color(values, cmap="seismic",vmin = 0,vmax = 1):
cNorm = colors.Normalize(vmin=vmin, vmax=vmax)
cMap = cmx.ScalarMappable(norm=cNorm, cmap=cmap)
return cMap.to_rgba(values,bytes=True)[:,:3]
def NP_Cosine(v1,v2):
d = np.sum(v1*v2,axis=1)
n1 = np.linalg.norm(v1,axis=1)
n2 = np.linalg.norm(v2,axis=1)
sim = (d/(n1*n2))
#print(sim.min())
return (1 - sim)/2
def log_scalar(name, scalar):
print(f"{name}:{scalar}")
def save_pc(filename, mesh):
save_dir=os.path.dirname(os.path.abspath(filename))
if not os.path.isdir(save_dir):
os.makedirs(save_dir)
test = mesh.visual.kind
#there is a bug in trimesh of it only saving normals when we tell the exporter explicitly to do so for point clouds.
#thus we are calling the exporter directly instead of mesh.export(...)
f = open(filename, "wb")
data = trimesh.exchange.ply.export_ply(mesh, 'binary',vertex_normal=True)
f.write(data)
f.close()
def Normalize(mesh):
coords = mesh.vertices
coord_max = np.max(coords, axis=0, keepdims=True)[0]
coord_min = np.min(coords, axis=0, keepdims=True)[0]
coord_center = 0.5*(coord_max + coord_min)
coords -= coord_center
scale = np.linalg.norm(coords,axis=1).max()
coords /= scale
def Compute_Chamfer(
mesheval : trimesh.Trimesh,
pointcloudgroundtruth : trimesh.PointCloud,
normalize = True,
normalize_base = True
) -> Dict[str,Union[float,trimesh.Trimesh]]:
normals = pointcloudgroundtruth.vertex_normals
if(normalize):
Normalize(mesheval)
Normalize(pointcloudgroundtruth)
elif(normalize_base):
Normalize(pointcloudgroundtruth)
#coords,idx = trimesh.sample.sample_surface_even(meshgt,numpoints)
#normals = meshgt.face_normals[idx,:]
coords = pointcloudgroundtruth.vertices
coords_new,idx_new = trimesh.sample.sample_surface_even(mesheval,coords.shape[0])
normals_new = mesheval.face_normals[idx_new,:]
print("Mean:")
print(coords_new.mean(axis=0), coords.mean(axis=0))
print(f"Using {coords.shape[0]} samples")
kdTree = KDTree(coords)
d1,idx1 = kdTree.query(coords_new)
#compute normal deformation
nd1 = NP_Cosine(normals[idx1,:],normals_new)
inverseTree = KDTree(coords_new)
d2,idx2 = inverseTree.query(coords)
#compute normal deformation
nd2 = NP_Cosine(normals_new[idx2,:],normals)
#mapping direction we either map
e2gt = "evalToGt"
gt2e = "gtToEval"
d1_max = d1.max()
d2_max = d2.max()
nd1_max = nd1.max()
nd2_max = nd2.max()
output = {}
output[e2gt] = d1.mean()
output[gt2e] = d2.mean()
output[f"{e2gt}_max"] = d1_max
output[f"{gt2e}_max"] = d2_max
output["hausdorff"] = max(d1_max,d2_max)
output["sum"] = (d1.mean()+d2.mean())
output[f"{e2gt}_normal"] = nd1.mean()
output[f"{gt2e}_normal"] = nd2.mean()
output["hausdorff_normal"] = max(nd1_max,nd2_max)
output["sum_normal"] = nd1.mean() + nd2.mean()
output[f"{gt2e}_pc"] = trimesh.Trimesh(vertices=coords_new,
vertex_colors=get_color(d1,"cool",d1.min(),d1.max()),
normals = normals_new
)
output[f"{e2gt}_pc"] = trimesh.Trimesh(vertices=coords,
vertex_colors=get_color(d2,"cool",d2.min(),d2.max()),
normals = normals
)
output[f"{gt2e}_pc_normal"] = trimesh.Trimesh(vertices=coords_new,
vertex_colors=get_color(nd1,"seismic",0,1),
normals = normals_new
)
output[f"{e2gt}_pc_normal"] = trimesh.Trimesh(vertices=coords,
vertex_colors=get_color(nd2,"seismic",0,1),
normals = normals)
return output
def Save_Chamfer(
mesheval : trimesh.Trimesh,
pointcloudgroundtruth : trimesh.PointCloud,
folder : str,
name : str = "chamfer_{0}_{1}",
save_mesh = True,
normalize = True,
normalize_base = True
):
output = Compute_Chamfer(mesheval,pointcloudgroundtruth,normalize,normalize_base)
for k,v in output:
if isinstance(v, float):
log_scalar(k,v)
elif save_mesh:
save_pc(f"{folder}/{k}.ply",v)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"--output",
required=True,
help="Folder for the ouput"
)
parser.add_argument(
"--input",
required=True,
help="""Format String for the input the first {0} will be replaced by the model name"""
)
parser.add_argument(
"--models",
default="data/benchmark_shapes/{0}_small_normalized.ply",
help="""Format string for the model folder the first {0} will be replace\
d by the name of the model"""
)
parser.add_argument(
"--normalize",
action="store_true",
help="""Normalize the meshes before evaluating"""
)
parser.add_argument(
"--normalize_base",
action="store_true",
help="""Normalize the meshes before evaluating"""
)
parser.add_argument(
"--samples",
default=4000000,type=int,
help="""Number of points to evaluate"""
)
original_stdout = sys.stdout # Save a reference to the original standard output
args = parser.parse_args()
input_wildcard = args.input
if( not os.path.isdir(args.output)):
os.makedirs(args.output)
std = Tee(f"{args.output}/chamfer.txt", 'w')
regex = re.compile(input_wildcard.format("([A-z_]*)"))
files = glob.glob(input_wildcard.format("[A-z_]*"))
files.sort()
for file in files:
print(file)
match = regex.search(file)
if(match):
if(len(match.groups())==0):
modelname = "static_model"
modelfile = args.models
else:
modelname = match.group(1)
modelfile = args.models.format(modelname)
if(os.path.exists(modelfile)):
print(f"\033[91m", file = original_stdout)
print(f"{modelname}:")
print(f"\033[0m", file = original_stdout)
meshEval = trimesh.load(file, process=False)
meshGt = load_point_cloud(modelfile,args.samples)
print(f"loaded {meshGt[0].shape[0]} points", file = original_stdout)
meshGt = trimesh.Trimesh(meshGt[0],vertex_normals=meshGt[1])
Save_Chamfer(meshEval,meshGt,f"{args.output}/{modelname}_", normalize=args.normalize, normalize_base=args.normalize_base)
del std
# hello