1 // Ceres Solver - A fast non-linear least squares minimizer
2 // Copyright 2013 Google Inc. All rights reserved.
3 // http://code.google.com/p/ceres-solver/
4 //
5 // Redistribution and use in source and binary forms, with or without
6 // modification, are permitted provided that the following conditions are met:
7 //
8 // * Redistributions of source code must retain the above copyright notice,
9 // this list of conditions and the following disclaimer.
10 // * Redistributions in binary form must reproduce the above copyright notice,
11 // this list of conditions and the following disclaimer in the documentation
12 // and/or other materials provided with the distribution.
13 // * Neither the name of Google Inc. nor the names of its contributors may be
14 // used to endorse or promote products derived from this software without
15 // specific prior written permission.
16 //
17 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
18 // AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 // ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
21 // LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
22 // CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
23 // SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
24 // INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
25 // CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
26 // ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
27 // POSSIBILITY OF SUCH DAMAGE.
28 //
29 // Author: mierle@gmail.com (Keir Mierle)
30
31 #include "ceres/c_api.h"
32
33 #include <cmath>
34
35 #include "glog/logging.h"
36 #include "gtest/gtest.h"
37
38 // Duplicated from curve_fitting.cc.
39 int num_observations = 67;
40 double data[] = {
41 0.000000e+00, 1.133898e+00,
42 7.500000e-02, 1.334902e+00,
43 1.500000e-01, 1.213546e+00,
44 2.250000e-01, 1.252016e+00,
45 3.000000e-01, 1.392265e+00,
46 3.750000e-01, 1.314458e+00,
47 4.500000e-01, 1.472541e+00,
48 5.250000e-01, 1.536218e+00,
49 6.000000e-01, 1.355679e+00,
50 6.750000e-01, 1.463566e+00,
51 7.500000e-01, 1.490201e+00,
52 8.250000e-01, 1.658699e+00,
53 9.000000e-01, 1.067574e+00,
54 9.750000e-01, 1.464629e+00,
55 1.050000e+00, 1.402653e+00,
56 1.125000e+00, 1.713141e+00,
57 1.200000e+00, 1.527021e+00,
58 1.275000e+00, 1.702632e+00,
59 1.350000e+00, 1.423899e+00,
60 1.425000e+00, 1.543078e+00,
61 1.500000e+00, 1.664015e+00,
62 1.575000e+00, 1.732484e+00,
63 1.650000e+00, 1.543296e+00,
64 1.725000e+00, 1.959523e+00,
65 1.800000e+00, 1.685132e+00,
66 1.875000e+00, 1.951791e+00,
67 1.950000e+00, 2.095346e+00,
68 2.025000e+00, 2.361460e+00,
69 2.100000e+00, 2.169119e+00,
70 2.175000e+00, 2.061745e+00,
71 2.250000e+00, 2.178641e+00,
72 2.325000e+00, 2.104346e+00,
73 2.400000e+00, 2.584470e+00,
74 2.475000e+00, 1.914158e+00,
75 2.550000e+00, 2.368375e+00,
76 2.625000e+00, 2.686125e+00,
77 2.700000e+00, 2.712395e+00,
78 2.775000e+00, 2.499511e+00,
79 2.850000e+00, 2.558897e+00,
80 2.925000e+00, 2.309154e+00,
81 3.000000e+00, 2.869503e+00,
82 3.075000e+00, 3.116645e+00,
83 3.150000e+00, 3.094907e+00,
84 3.225000e+00, 2.471759e+00,
85 3.300000e+00, 3.017131e+00,
86 3.375000e+00, 3.232381e+00,
87 3.450000e+00, 2.944596e+00,
88 3.525000e+00, 3.385343e+00,
89 3.600000e+00, 3.199826e+00,
90 3.675000e+00, 3.423039e+00,
91 3.750000e+00, 3.621552e+00,
92 3.825000e+00, 3.559255e+00,
93 3.900000e+00, 3.530713e+00,
94 3.975000e+00, 3.561766e+00,
95 4.050000e+00, 3.544574e+00,
96 4.125000e+00, 3.867945e+00,
97 4.200000e+00, 4.049776e+00,
98 4.275000e+00, 3.885601e+00,
99 4.350000e+00, 4.110505e+00,
100 4.425000e+00, 4.345320e+00,
101 4.500000e+00, 4.161241e+00,
102 4.575000e+00, 4.363407e+00,
103 4.650000e+00, 4.161576e+00,
104 4.725000e+00, 4.619728e+00,
105 4.800000e+00, 4.737410e+00,
106 4.875000e+00, 4.727863e+00,
107 4.950000e+00, 4.669206e+00,
108 };
109
110 // A test cost function, similar to the one in curve_fitting.c.
exponential_residual(void * user_data,double ** parameters,double * residuals,double ** jacobians)111 int exponential_residual(void* user_data,
112 double** parameters,
113 double* residuals,
114 double** jacobians) {
115 double* measurement = (double*) user_data;
116 double x = measurement[0];
117 double y = measurement[1];
118 double m = parameters[0][0];
119 double c = parameters[1][0];
120
121 residuals[0] = y - exp(m * x + c);
122 if (jacobians == NULL) {
123 return 1;
124 }
125 if (jacobians[0] != NULL) {
126 jacobians[0][0] = - x * exp(m * x + c); // dr/dm
127 }
128 if (jacobians[1] != NULL) {
129 jacobians[1][0] = - exp(m * x + c); // dr/dc
130 }
131 return 1;
132 }
133
134 namespace ceres {
135 namespace internal {
136
TEST(C_API,SimpleEndToEndTest)137 TEST(C_API, SimpleEndToEndTest) {
138 double m = 0.0;
139 double c = 0.0;
140 double *parameter_pointers[] = { &m, &c };
141 int parameter_sizes[] = { 1, 1 };
142
143 ceres_problem_t* problem = ceres_create_problem();
144 for (int i = 0; i < num_observations; ++i) {
145 ceres_problem_add_residual_block(
146 problem,
147 exponential_residual, // Cost function
148 &data[2 * i], // Points to the (x,y) measurement
149 NULL, // Loss function
150 NULL, // Loss function user data
151 1, // Number of residuals
152 2, // Number of parameter blocks
153 parameter_sizes,
154 parameter_pointers);
155 }
156
157 ceres_solve(problem);
158
159 EXPECT_NEAR(0.3, m, 0.02);
160 EXPECT_NEAR(0.1, c, 0.04);
161
162 ceres_free_problem(problem);
163 }
164
165 template<typename T>
166 class ScopedSetValue {
167 public:
ScopedSetValue(T * variable,T new_value)168 ScopedSetValue(T* variable, T new_value)
169 : variable_(variable), old_value_(*variable) {
170 *variable = new_value;
171 }
~ScopedSetValue()172 ~ScopedSetValue() {
173 *variable_ = old_value_;
174 }
175
176 private:
177 T* variable_;
178 T old_value_;
179 };
180
TEST(C_API,LossFunctions)181 TEST(C_API, LossFunctions) {
182 double m = 0.2;
183 double c = 0.03;
184 double *parameter_pointers[] = { &m, &c };
185 int parameter_sizes[] = { 1, 1 };
186
187 // Create two outliers, but be careful to leave the data intact.
188 ScopedSetValue<double> outlier1x(&data[12], 2.5);
189 ScopedSetValue<double> outlier1y(&data[13], 1.0e3);
190 ScopedSetValue<double> outlier2x(&data[14], 3.2);
191 ScopedSetValue<double> outlier2y(&data[15], 30e3);
192
193 // Create a cauchy cost function, and reuse it many times.
194 void* cauchy_loss_data =
195 ceres_create_cauchy_loss_function_data(5.0);
196
197 ceres_problem_t* problem = ceres_create_problem();
198 for (int i = 0; i < num_observations; ++i) {
199 ceres_problem_add_residual_block(
200 problem,
201 exponential_residual, // Cost function
202 &data[2 * i], // Points to the (x,y) measurement
203 ceres_stock_loss_function,
204 cauchy_loss_data, // Loss function user data
205 1, // Number of residuals
206 2, // Number of parameter blocks
207 parameter_sizes,
208 parameter_pointers);
209 }
210
211 ceres_solve(problem);
212
213 EXPECT_NEAR(0.3, m, 0.02);
214 EXPECT_NEAR(0.1, c, 0.04);
215
216 ceres_free_stock_loss_function_data(cauchy_loss_data);
217 ceres_free_problem(problem);
218 }
219
220 } // namespace internal
221 } // namespace ceres
222