-
Notifications
You must be signed in to change notification settings - Fork 16
/
utils.py
196 lines (165 loc) · 6.12 KB
/
utils.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
# --------------------------------------------------------
# Dual Octree Graph Networks
# Copyright (c) 2022 Microsoft
# Licensed under The MIT License [see LICENSE for details]
# Written by Peng-Shuai Wang
# --------------------------------------------------------
# autopep8: off
import ocnn
import torch
import torch.autograd
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import skimage.measure
import trimesh
from plyfile import PlyData, PlyElement
from scipy.spatial import cKDTree
# autopep8: on
def get_mgrid(size, dim=3):
r'''
Example:
>>> get_mgrid(3, dim=2)
array([[0.0, 0.0],
[0.0, 1.0],
[0.0, 2.0],
[1.0, 0.0],
[1.0, 1.0],
[1.0, 2.0],
[2.0, 0.0],
[2.0, 1.0],
[2.0, 2.0]], dtype=float32)
'''
coord = np.arange(0, size, dtype=np.float32)
coords = [coord] * dim
output = np.meshgrid(*coords, indexing='ij')
output = np.stack(output, -1)
output = output.reshape(size**dim, dim)
return output
def lin2img(tensor):
channels = 1
num_samples = tensor.shape
size = int(np.sqrt(num_samples))
return tensor.view(channels, size, size)
def make_contour_plot(array_2d, mode='log'):
fig, ax = plt.subplots(figsize=(2.75, 2.75), dpi=300)
if(mode == 'log'):
nlevels = 6
levels_pos = np.logspace(-2, 0, num=nlevels) # logspace
levels_neg = -1. * levels_pos[::-1]
levels = np.concatenate((levels_neg, np.zeros((0)), levels_pos), axis=0)
colors = plt.get_cmap("Spectral")(np.linspace(0., 1., num=nlevels * 2 + 1))
elif(mode == 'lin'):
nlevels = 10
levels = np.linspace(-.5, .5, num=nlevels)
colors = plt.get_cmap("Spectral")(np.linspace(0., 1., num=nlevels))
else:
raise NotImplementedError
sample = np.flipud(array_2d)
CS = ax.contourf(sample, levels=levels, colors=colors)
cbar = fig.colorbar(CS)
ax.contour(sample, levels=levels, colors='k', linewidths=0.1)
ax.contour(sample, levels=[0], colors='k', linewidths=0.3)
ax.axis('off')
return fig
def write_sdf_summary(model, writer, global_step, alias=''):
size = 128
coords_2d = get_mgrid(size, dim=2)
coords_2d = coords_2d / size - 1.0 # [0, size] -> [-1, 1]
coords_2d = torch.from_numpy(coords_2d)
with torch.no_grad():
zeros = torch.zeros_like(coords_2d[:, :1])
ones = torch.ones_like(coords_2d[:, :1])
names = ['train_yz_sdf_slice', 'train_xz_sdf_slice', 'train_xy_sdf_slice']
coords = [torch.cat((zeros, coords_2d), dim=-1),
torch.cat((coords_2d[:, :1], zeros, coords_2d[:, -1:]), dim=-1),
torch.cat((coords_2d, -0.75 * ones), dim=-1)]
for name, coord in zip(names, coords):
ids = torch.zeros(coord.shape[0], 1)
coord = torch.cat([coord, ids], dim=1).cuda()
sdf_values = model(coord)
sdf_values = lin2img(sdf_values).squeeze().cpu().numpy()
fig = make_contour_plot(sdf_values)
writer.add_figure(alias + name, fig, global_step=global_step)
def calc_sdf(model, size=256, max_batch=64**3, bbmin=-1.0, bbmax=1.0):
# generate samples
num_samples = size ** 3
samples = get_mgrid(size, dim=3)
samples = samples * ((bbmax - bbmin) / size) + bbmin # [0,sz]->[bbmin,bbmax]
samples = torch.from_numpy(samples)
sdfs = torch.zeros(num_samples)
# forward
head = 0
while head < num_samples:
tail = min(head + max_batch, num_samples)
sample_subset = samples[head:tail, :]
idx = torch.zeros(sample_subset.shape[0], 1)
pts = torch.cat([sample_subset, idx], dim=1).cuda()
pred = model(pts).squeeze().detach().cpu()
sdfs[head:tail] = pred
head += max_batch
sdfs = sdfs.reshape(size, size, size).numpy()
return sdfs
def create_mesh(model, filename, size=256, max_batch=64**3, level=0,
bbmin=-0.9, bbmax=0.9, mesh_scale=1.0, save_sdf=False, **kwargs):
# marching cubes
sdf_values = calc_sdf(model, size, max_batch, bbmin, bbmax)
vtx, faces = np.zeros((0, 3)), np.zeros((0, 3))
try:
vtx, faces, _, _ = skimage.measure.marching_cubes(sdf_values, level)
except:
pass
if vtx.size == 0 or faces.size == 0:
print('Warning from marching cubes: Empty mesh!')
return
# normalize vtx
vtx = vtx * ((bbmax - bbmin) / size) + bbmin # [0,sz]->[bbmin,bbmax]
vtx = vtx * mesh_scale # rescale
# save to ply and npy
mesh = trimesh.Trimesh(vtx, faces)
mesh.export(filename)
if save_sdf:
np.save(filename[:-4] + ".sdf.npy", sdf_values)
def calc_sdf_err(filename_gt, filename_pred):
scale = 1.0e2 # scale the result for better display
sdf_gt = np.load(filename_gt)
sdf = np.load(filename_pred)
err = np.abs(sdf - sdf_gt).mean() * scale
return err
def calc_chamfer(filename_gt, filename_pred, point_num):
scale = 1.0e5 # scale the result for better display
np.random.seed(101)
mesh_a = trimesh.load(filename_gt)
points_a, _ = trimesh.sample.sample_surface(mesh_a, point_num)
mesh_b = trimesh.load(filename_pred)
points_b, _ = trimesh.sample.sample_surface(mesh_b, point_num)
kdtree_a = cKDTree(points_a)
dist_a, _ = kdtree_a.query(points_b)
chamfer_a = np.mean(np.square(dist_a)) * scale
kdtree_b = cKDTree(points_b)
dist_b, _ = kdtree_b.query(points_a)
chamfer_b = np.mean(np.square(dist_b)) * scale
return chamfer_a, chamfer_b
def points2ply(filename, points, scale=1.0):
xyz = ocnn.points_property(points, 'xyz')
normal = ocnn.points_property(points, 'normal')
has_normal = normal is not None
xyz = xyz.numpy() * scale
if has_normal: normal = normal.numpy()
# data types
data = xyz
py_types = (float, float, float)
npy_types = [('x', 'f4'), ('y', 'f4'), ('z', 'f4')]
if has_normal:
py_types = py_types + (float, float, float)
npy_types = npy_types + [('nx', 'f4'), ('ny', 'f4'), ('nz', 'f4')]
data = np.concatenate((data, normal), axis=1)
# format into NumPy structured array
vertices = []
for idx in range(data.shape[0]):
vertices.append(tuple(dtype(d) for dtype, d in zip(py_types, data[idx])))
structured_array = np.array(vertices, dtype=npy_types)
el = PlyElement.describe(structured_array, 'vertex')
# write ply
PlyData([el]).write(filename)