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
|
# ----------------------------------------------------------------------------
# - Open3D: www.open3d.org -
# ----------------------------------------------------------------------------
# The MIT License (MIT)
#
# Copyright (c) 2018-2021 www.open3d.org
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
# IN THE SOFTWARE.
# ----------------------------------------------------------------------------
import open3d as o3d
import numpy as np
import pytest
import mltest
from check_gradients import check_gradients
# skip all tests if the ml ops were not built
pytestmark = mltest.default_marks
# the supported dtypes
position_dtypes = pytest.mark.parametrize('pos_dtype', [np.float32, np.float64])
feature_dtypes = pytest.mark.parametrize(
'feat_dtype', [np.float32, np.float64, np.int32, np.int64])
# aggregation functions
position_functions = pytest.mark.parametrize(
'position_fn', ['average', 'center', 'nearest_neighbor'])
feature_functions = pytest.mark.parametrize(
'feature_fn', ['average', 'max', 'nearest_neighbor'])
@mltest.parametrize.ml_cpu_only
@position_dtypes
@feature_dtypes
@position_functions
@feature_functions
def test_voxel_pooling(ml, pos_dtype, feat_dtype, position_fn, feature_fn):
# yapf: disable
points = np.array([
# 3 points in voxel
[0.5, 0.5, 0.5],
[0.7, 0.2, 0.3],
[0.7, 0.5, 0.9],
# 2 points in another voxel
[1.4, 1.5, 1.4],
[1.7, 1.2, 1.3],
], dtype=pos_dtype)
features = np.array([
# 3 points in voxel
[1,1],
[2,1],
[3,1],
# 2 points in another voxel
[4,1],
[5,1],
], dtype=feat_dtype)
# yapf: enable
voxel_size = 1
ans = mltest.run_op(ml, ml.device, True, ml.ops.voxel_pooling, points,
features, voxel_size, position_fn, feature_fn)
if position_fn == 'average':
expected_positions = np.stack(
[np.mean(points[:3], axis=0),
np.mean(points[3:], axis=0)])
elif position_fn == 'center':
expected_positions = np.array([[0.5, 0.5, 0.5], [1.5, 1.5, 1.5]],
dtype=pos_dtype)
elif position_fn == 'nearest_neighbor':
expected_positions = np.array([points[0], points[3]], dtype=pos_dtype)
assert len(ans.pooled_positions) == 2
# compute assignment
if np.linalg.norm(ans.pooled_positions[0] -
expected_positions[0]) < np.linalg.norm(
ans.pooled_positions[0] - expected_positions[1]):
index = [0, 1]
else:
index = [1, 0]
np.testing.assert_allclose(ans.pooled_positions, expected_positions[index])
if feature_fn == 'average':
if np.issubdtype(feat_dtype, np.integer):
expected_features = np.stack([
np.sum(features[:3], axis=0) // 3,
np.sum(features[3:], axis=0) // 2
])
else:
expected_features = np.stack(
[np.mean(features[:3], axis=0),
np.mean(features[3:], axis=0)])
elif feature_fn == 'max':
expected_features = np.stack(
[np.max(features[:3], axis=0),
np.max(features[3:], axis=0)])
elif feature_fn == 'nearest_neighbor':
expected_features = np.array([features[0], features[3]])
np.testing.assert_allclose(ans.pooled_features, expected_features[index])
@mltest.parametrize.ml_cpu_only
@position_dtypes
@feature_dtypes
@position_functions
@feature_functions
def test_voxel_pooling_empty_point_set(ml, pos_dtype, feat_dtype, position_fn,
feature_fn):
points = np.zeros(shape=[0, 3], dtype=pos_dtype)
features = np.zeros(shape=[0, 5], dtype=feat_dtype)
voxel_size = 1
ans = mltest.run_op(ml, ml.device, True, ml.ops.voxel_pooling, points,
features, voxel_size, position_fn, feature_fn)
np.testing.assert_array_equal(points, ans.pooled_positions)
np.testing.assert_array_equal(features, ans.pooled_features)
# tf and torch does not support gradient computation for integer types
gradient_feature_dtypes = pytest.mark.parametrize('feat_dtype',
[np.float32, np.float64])
@mltest.parametrize.ml_cpu_only
@position_dtypes
@gradient_feature_dtypes
@position_functions
@feature_functions
@pytest.mark.parametrize('empty_point_set', [
False,
])
def test_voxel_pooling_grad(ml, pos_dtype, feat_dtype, position_fn, feature_fn,
empty_point_set):
rng = np.random.RandomState(123)
N = 0 if empty_point_set else 50
channels = 4
positions = rng.uniform(0, 1, (N, 3)).astype(pos_dtype)
# generate features and make sure that the feature values are not too close to each other
# if they are too close the numerical jacobian will be wrong
features = np.linspace(0, N * channels, num=N * channels, endpoint=False)
np.random.shuffle(features)
features = np.reshape(features, (N, channels)).astype(feat_dtype)
voxel_size = 0.25
def fn(features):
ans = mltest.run_op(ml, ml.device, True, ml.ops.voxel_pooling,
positions, features, voxel_size, position_fn,
feature_fn)
return ans.pooled_features
def fn_grad(features_bp, features):
return mltest.run_op_grad(ml, ml.device, True, ml.ops.voxel_pooling,
features, 'pooled_features', features_bp,
positions, features, voxel_size, position_fn,
feature_fn)
gradient_OK = check_gradients(features, fn, fn_grad, epsilon=1)
assert gradient_OK
|
