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
|
// ----------------------------------------------------------------------------
// - Open3D: www.open3d.org -
// ----------------------------------------------------------------------------
// Copyright (c) 2018-2024 www.open3d.org
// SPDX-License-Identifier: MIT
// ----------------------------------------------------------------------------
#include "open3d/pipelines/odometry/Odometry.h"
#include "open3d/geometry/Image.h"
#include "open3d/geometry/RGBDImage.h"
#include "open3d/pipelines/odometry/OdometryOption.h"
#include "open3d/pipelines/odometry/RGBDOdometryJacobian.h"
#include "pybind/docstring.h"
#include "pybind/pipelines/odometry/odometry.h"
namespace open3d {
namespace pipelines {
namespace odometry {
template <class RGBDOdometryJacobianBase = RGBDOdometryJacobian>
class PyRGBDOdometryJacobian : public RGBDOdometryJacobianBase {
public:
using RGBDOdometryJacobianBase::RGBDOdometryJacobianBase;
void ComputeJacobianAndResidual(
int row,
std::vector<Eigen::Vector6d, utility::Vector6d_allocator> &J_r,
std::vector<double> &r,
std::vector<double> &w,
const geometry::RGBDImage &source,
const geometry::RGBDImage &target,
const geometry::Image &source_xyz,
const geometry::RGBDImage &target_dx,
const geometry::RGBDImage &target_dy,
const Eigen::Matrix3d &intrinsic,
const Eigen::Matrix4d &extrinsic,
const CorrespondenceSetPixelWise &corresps) const override {
PYBIND11_OVERLOAD_PURE(void, RGBDOdometryJacobianBase, row, J_r, r,
source, target, source_xyz, target_dx, target_dy,
extrinsic, corresps, intrinsic);
}
};
void pybind_odometry_declarations(py::module &m) {
py::module m_odometry = m.def_submodule("odometry", "Odometry pipeline.");
py::class_<OdometryOption> odometry_option(
m_odometry, "OdometryOption",
"Class that defines Odometry options.");
py::class_<RGBDOdometryJacobian,
PyRGBDOdometryJacobian<RGBDOdometryJacobian>>
jacobian(
m_odometry, "RGBDOdometryJacobian",
"Base class that computes Jacobian from two RGB-D images.");
py::class_<RGBDOdometryJacobianFromColorTerm,
PyRGBDOdometryJacobian<RGBDOdometryJacobianFromColorTerm>,
RGBDOdometryJacobian>
jacobian_color(m_odometry, "RGBDOdometryJacobianFromColorTerm",
R"(Class to Compute Jacobian using color term.
Energy: :math:`(I_p-I_q)^2.`
Reference:
F. Steinbrucker, J. Sturm, and D. Cremers.
Real-time visual odometry from dense RGB-D images.
In ICCV Workshops, 2011.)");
py::class_<RGBDOdometryJacobianFromHybridTerm,
PyRGBDOdometryJacobian<RGBDOdometryJacobianFromHybridTerm>,
RGBDOdometryJacobian>
jacobian_hybrid(m_odometry, "RGBDOdometryJacobianFromHybridTerm",
R"(Class to compute Jacobian using hybrid term
Energy: :math:`(I_p-I_q)^2 + \lambda(D_p-D_q')^2`
Reference:
J. Park, Q.-Y. Zhou, and V. Koltun
Anonymous submission.)");
}
void pybind_odometry_definitions(py::module &m) {
auto m_odometry = static_cast<py::module>(m.attr("odometry"));
// open3d.odometry.OdometryOption
auto odometry_option = static_cast<py::class_<OdometryOption>>(
m_odometry.attr("OdometryOption"));
odometry_option
.def(py::init(
[](std::vector<int> iteration_number_per_pyramid_level,
double depth_diff_max, double depth_min,
double depth_max) {
return new OdometryOption(
iteration_number_per_pyramid_level,
depth_diff_max, depth_min, depth_max);
}),
"iteration_number_per_pyramid_level"_a =
std::vector<int>{20, 10, 5},
"depth_diff_max"_a = 0.03, "depth_min"_a = 0.0,
"depth_max"_a = 4.0)
.def_readwrite("iteration_number_per_pyramid_level",
&OdometryOption::iteration_number_per_pyramid_level_,
"List(int): Iteration number per image pyramid "
"level, typically larger image in the pyramid have "
"lower iteration number to reduce computation "
"time.")
.def_readwrite("depth_diff_max", &OdometryOption::depth_diff_max_,
"Maximum depth difference to be considered as "
"correspondence. In depth image domain, if two "
"aligned pixels have a depth difference less than "
"specified value, they are considered as a "
"correspondence. Larger value induce more "
"aggressive search, but it is prone to unstable "
"result.")
.def_readwrite("depth_min", &OdometryOption::depth_min_,
"Pixels that has smaller than specified depth "
"values are ignored.")
.def_readwrite("depth_max", &OdometryOption::depth_max_,
"Pixels that has larger than specified depth values "
"are ignored.")
.def("__repr__", [](const OdometryOption &c) {
int num_pyramid_level =
(int)c.iteration_number_per_pyramid_level_.size();
std::string str_iteration_number_per_pyramid_level_ = "[ ";
for (int i = 0; i < num_pyramid_level; i++)
str_iteration_number_per_pyramid_level_ +=
std::to_string(
c.iteration_number_per_pyramid_level_[i]) +
", ";
str_iteration_number_per_pyramid_level_ += "] ";
return fmt::format(
"OdometryOption(\n"
"iteration_number_per_pyramid_level={},\n"
"depth_diff_max={},\n"
"depth_min={},\n"
"depth_max={},\n"
")",
str_iteration_number_per_pyramid_level_,
c.depth_diff_max_, c.depth_min_, c.depth_max_);
});
// open3d.odometry.RGBDOdometryJacobian
auto jacobian = static_cast<
py::class_<RGBDOdometryJacobian,
PyRGBDOdometryJacobian<RGBDOdometryJacobian>>>(
m_odometry.attr("RGBDOdometryJacobian"));
jacobian.def(
"compute_jacobian_and_residual",
&RGBDOdometryJacobian::ComputeJacobianAndResidual,
py::call_guard<py::gil_scoped_release>(),
"Function to compute i-th row of J and r the vector form of J_r is "
"basically 6x1 matrix, but it can be easily extendable to 6xn "
"matrix. See RGBDOdometryJacobianFromHybridTerm for this case."
"row"_a,
"J_r"_a, "r"_a, "w"_a, "source"_a, "target"_a, "source_xyz"_a,
"target_dx"_a, "target_dy"_a, "intrinsic"_a, "extrinsic"_a,
"corresps"_a);
// open3d.odometry.RGBDOdometryJacobianFromColorTerm: RGBDOdometryJacobian
auto jacobian_color = static_cast<py::class_<
RGBDOdometryJacobianFromColorTerm,
PyRGBDOdometryJacobian<RGBDOdometryJacobianFromColorTerm>,
RGBDOdometryJacobian>>(
m_odometry.attr("RGBDOdometryJacobianFromColorTerm"));
py::detail::bind_default_constructor<RGBDOdometryJacobianFromColorTerm>(
jacobian_color);
py::detail::bind_copy_functions<RGBDOdometryJacobianFromColorTerm>(
jacobian_color);
jacobian_color.def(
"__repr__", [](const RGBDOdometryJacobianFromColorTerm &te) {
return std::string("RGBDOdometryJacobianFromColorTerm");
});
// open3d.odometry.RGBDOdometryJacobianFromHybridTerm: RGBDOdometryJacobian
auto jacobian_hybrid = static_cast<py::class_<
RGBDOdometryJacobianFromHybridTerm,
PyRGBDOdometryJacobian<RGBDOdometryJacobianFromHybridTerm>,
RGBDOdometryJacobian>>(
m_odometry.attr("RGBDOdometryJacobianFromHybridTerm"));
py::detail::bind_default_constructor<RGBDOdometryJacobianFromHybridTerm>(
jacobian_hybrid);
py::detail::bind_copy_functions<RGBDOdometryJacobianFromHybridTerm>(
jacobian_hybrid);
jacobian_hybrid.def(
"__repr__", [](const RGBDOdometryJacobianFromHybridTerm &te) {
return std::string("RGBDOdometryJacobianFromHybridTerm");
});
m_odometry.def(
"compute_rgbd_odometry", &ComputeRGBDOdometry,
py::call_guard<py::gil_scoped_release>(),
"Function to estimate 6D rigid motion from two RGBD image pairs. "
"Output: (is_success, 4x4 motion matrix, 6x6 information matrix).",
"rgbd_source"_a, "rgbd_target"_a,
"pinhole_camera_intrinsic"_a = camera::PinholeCameraIntrinsic(),
"odo_init"_a = Eigen::Matrix4d::Identity(),
"jacobian"_a = RGBDOdometryJacobianFromHybridTerm(),
"option"_a = OdometryOption());
docstring::FunctionDocInject(
m_odometry, "compute_rgbd_odometry",
{
{"rgbd_source", "Source RGBD image."},
{"rgbd_target", "Target RGBD image."},
{"pinhole_camera_intrinsic", "Camera intrinsic parameters"},
{"odo_init", "Initial 4x4 motion matrix estimation."},
{"jacobian",
"The odometry Jacobian method to use. Can be "
"``"
"RGBDOdometryJacobianFromHybridTerm()`` or "
"``RGBDOdometryJacobianFromColorTerm("
").``"},
{"option", "Odometry hyper parameters."},
});
m_odometry.def("compute_correspondence", &ComputeCorrespondence,
py::call_guard<py::gil_scoped_release>(),
"Function to estimate point to point correspondences from "
"two depth "
"images. A vector of u_s, v_s, u_t, v_t which maps the 2d "
"coordinates of source to target.",
"intrinsic_matrix"_a, "extrinsic"_a, "depth_s"_a,
"depth_t"_a, "option"_a = OdometryOption());
docstring::FunctionDocInject(
m_odometry, "compute_correspondence",
{
{"intrinsic_matrix", "Camera intrinsic parameters."},
{"extrinsic",
"Estimation of transform from source to target."},
{"depth_s", "Source depth image."},
{"depth_t", "Target depth image."},
{"option", "Odometry hyper parameters."},
});
}
} // namespace odometry
} // namespace pipelines
} // namespace open3d
|
