-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathscore_and_vis.py
207 lines (191 loc) · 8.46 KB
/
score_and_vis.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
import argparse
import os
import numpy as np
import scipy.spatial as spatial
from tqdm import tqdm
import plotly.graph_objects as go
def parse_arguments():
parser = argparse.ArgumentParser()
parser.add_argument('--indir', type=str, default='./pclouds', help='input folder (point clouds)')
parser.add_argument('--filename', type=str, default='noise_free_and_noise.txt',
help='output folder (trained models)')
parser.add_argument('--logdir', type=str, default='./logs', help='training log folder')
parser.add_argument('--html_dir', type=str, default='./htmlfiles', help='htmlfiles folder')
parser.add_argument('--num_neighbours', type=int, default=10, help='num_neighbours')
parser.add_argument('--points_per_patch', type=int, default=500, help='max. number of points per patch')
return parser.parse_args()
def compute_ang_err(n0, n):
n0 = np.broadcast_to(np.expand_dims(n0, axis=0), (len(n), 3))
err = np.minimum(((n0 - n) ** 2).sum(1), ((n0 + n) ** 2).sum(1))
return err
def visualise(opt, plot_index, neigh_pts, neigh_normals, pts, normal0):
arrow_tip_ratio = 0.2
arrow_starting_ratio = 0.98
normal0 = normal0 / 5
neigh_normals = neigh_normals / 5
if not os.path.exists(opt.html_dir):
os.mkdir(opt.html_dir)
html_path = os.path.join(opt.html_dir, "%d.html" % plot_index)
layout = go.Layout(title='3D Gaussian Normal Visualization',
scene=dict(aspectmode='cube', # this string can be 'data', 'cube', 'auto', 'manual'
# a custom aspectratio is defined as follows:
xaxis=dict(
# range=[-1, 1],
showgrid=False, # thin lines in the background
zeroline=False, # thick line at x=0
visible=False),
yaxis=dict(
# range=[-1, 1],
showgrid=False, # thin lines in the background
zeroline=False, # thick line at x=0
visible=False
),
zaxis=dict(
# range=[-1, 1],
showgrid=False, # thin lines in the background
zeroline=False, # thick line at x=0
visible=False
),
aspectratio=dict(x=1, y=1, z=1)
)
)
fig = go.Figure(layout=layout)
# for i, point in enumerate(pts):
# fig.add_trace(go.Scatter3d(
# x=[point[0], point[0] + normal[i, 0] / 2],
# y=[point[1], point[1] + normal[i, 1] / 2],
# z=[point[2], point[2] + normal[i, 2] / 2],
# mode='lines',
# showlegend=False,
# line=dict(width=2, color='rgb(0, 255, 0)')
# ))
for i, point in enumerate(neigh_pts):
fig.add_trace(go.Scatter3d(
x=[point[0], point[0] + neigh_normals[i, 0] / 2],
y=[point[1], point[1] + neigh_normals[i, 1] / 2],
z=[point[2], point[2] + neigh_normals[i, 2] / 2],
mode='lines',
showlegend=False,
line=dict(width=2, color='rgb(0, 255, 0)')
))
fig.add_trace(go.Cone(
x=[point[0] + arrow_starting_ratio * neigh_normals[i, 0] / 2],
y=[point[1] + arrow_starting_ratio * neigh_normals[i, 1] / 2],
z=[point[2] + arrow_starting_ratio * neigh_normals[i, 2] / 2],
u=[point[0] + arrow_tip_ratio * neigh_normals[i, 0] / 2],
v=[point[1] + arrow_tip_ratio * neigh_normals[i, 1] / 2],
w=[point[2] + arrow_tip_ratio * neigh_normals[i, 2] / 2],
showlegend=False,
showscale=False,
colorscale=[[0, 'rgb(255,0,0)'], [1, 'rgb(255,0,0)']]
))
fig.add_trace(go.Scatter3d(
x=neigh_pts[:, 0],
y=neigh_pts[:, 1],
z=neigh_pts[:, 2],
mode='markers',
showlegend=False,
marker=dict(
size=1,
color='blue',
),
))
fig.add_trace(go.Scatter3d(
x=pts[:, 0],
y=pts[:, 1],
z=pts[:, 2],
mode='markers',
showlegend=False,
marker=dict(
size=1,
color='blue',
),
))
fig.add_trace(go.Scatter3d(
x=np.array([0]),
y=np.array([0]),
z=np.array([0]),
mode='markers',
showlegend=False,
marker=dict(
size=1,
color='red',
),
))
fig.add_trace(go.Scatter3d(
x=[0, normal0[0]],
y=[0, normal0[1]],
z=[0, normal0[2]],
mode='lines',
showlegend=False,
line=dict(width=2, color='black')
))
fig.add_trace(go.Cone(
x=[arrow_starting_ratio * normal0[0]],
y=[arrow_starting_ratio * normal0[1]],
z=[arrow_starting_ratio * normal0[2]],
u=[arrow_tip_ratio * normal0[0]],
v=[arrow_tip_ratio * normal0[1]],
w=[arrow_tip_ratio * normal0[2]],
showlegend=False,
showscale=False,
colorscale=[[0, 'rgb(255,0,0)'], [1, 'rgb(255,0,0)']]
))
# fig.show()
fig.write_html(html_path)
print('html file has been saved to %s ' % html_path)
class PointCloud:
def __init__(self, opt, type='noise_free'):
self.root = opt.indir
self.shape_names = []
self.patch_size = opt.points_per_patch
self.k = opt.num_neighbours
self.seed = np.random.randint(0, 2 ** 32)
self.rng = np.random.RandomState(self.seed)
with open(os.path.join(self.root, opt.filename)) as f:
self.shape_names = f.readlines()
self.shape_names = [x.strip() for x in self.shape_names]
self.shape_names = list(filter(None, self.shape_names))
if type == 'noise_free':
point_filename = os.path.join(self.root, self.shape_names[0] + '.xyz')
normals_filename = os.path.join(self.root, self.shape_names[0] + '.normals')
elif type == 'noise':
point_filename = os.path.join(self.root, self.shape_names[1] + '.xyz')
normals_filename = os.path.join(self.root, self.shape_names[1] + '.normals')
else:
raise ValueError('Unknown type of PointCloud: noise_free ot noise ?')
self.pts = np.loadtxt(point_filename)
self.normals = np.loadtxt(normals_filename)
def __len__(self):
return len(self.pts)
def compute_score(self, noise_free_pcloud):
print("Computing the scores !".center(50, "-"))
iim = np.empty((0, self.k))
ssm = []
noise_free_pcloud.nftree = spatial.cKDTree(noise_free_pcloud.pts, 500)
for j, point in enumerate(tqdm(self.pts)):
dd, ii = noise_free_pcloud.nftree.query(point, k=self.k)
iim = np.append(iim, np.expand_dims(ii, axis=0), axis=0)
ang_err = compute_ang_err(self.normals[j], noise_free_pcloud.normals[ii])
score = np.sum(ang_err / dd)
ssm.append(score)
return np.array(ssm), iim
def get_patch(self, center_point):
# _, patch_inds = self.nftree.query(center_point, self.patch_size)
patch_inds = self.rng.choice(self.__len__(), int(0.1 * self.__len__()), replace=False)
patch_normals = self.normals[patch_inds]
patch_pts = self.pts[patch_inds] - center_point
patch_rad = float(np.linalg.norm(patch_pts.max(0) - patch_pts.min(0), 2))
patch_pts = patch_pts / patch_rad
return patch_pts, patch_normals, patch_rad
if __name__ == '__main__':
opt = parse_arguments()
nfp = PointCloud(opt, type='noise_free')
noisep = PointCloud(opt, type='noise')
score, iim = noisep.compute_score(nfp)
sort_index = np.argsort(score)[::-1]
plot_index = int(input('Please enter a number you wanna plot from 1 to %d :' % len(score)))
patch_pts, patch_normals, rad = nfp.get_patch(noisep.pts[sort_index[plot_index - 1]])
neigh_pts = (nfp.pts[iim[sort_index[plot_index - 1]].astype(int)] - noisep.pts[sort_index[plot_index - 1]]) / rad
neigh_normals = nfp.normals[iim[sort_index[plot_index - 1]].astype(int)]
visualise(opt, plot_index, neigh_pts, neigh_normals, patch_pts, noisep.normals[sort_index[plot_index - 1]])