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
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
|
// Copyright (c) 2021, Viktor Larsson
// All rights reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
//
// * Redistributions in binary form must reproduce the above copyright
// notice, this list of conditions and the following disclaimer in the
// documentation and/or other materials provided with the distribution.
//
// * Neither the name of the copyright holder nor the
// names of its contributors may be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
// ARE DISCLAIMED. IN NO EVENT SHALL COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE
// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
// (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
// LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
// ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#if __GNUC__ >= 12
#pragma GCC diagnostic ignored "-Warray-bounds"
#endif
#include "bundle.h"
#include "PoseLib/robust/jacobian_impl.h"
#include "PoseLib/robust/lm_impl.h"
#include "PoseLib/robust/robust_loss.h"
#include <iostream>
namespace poselib {
////////////////////////////////////////////////////////////////////////
// Below here we have wrappers for the refinement
// These are super messy due to the loss functions being templated
// and the hack we use to handle weights
// (see UniformWeightVector in jacobian_impl.h)
#define SWITCH_LOSS_FUNCTIONS \
case BundleOptions::LossType::TRIVIAL: \
SWITCH_LOSS_FUNCTION_CASE(TrivialLoss); \
break; \
case BundleOptions::LossType::TRUNCATED: \
SWITCH_LOSS_FUNCTION_CASE(TruncatedLoss); \
break; \
case BundleOptions::LossType::HUBER: \
SWITCH_LOSS_FUNCTION_CASE(HuberLoss); \
break; \
case BundleOptions::LossType::CAUCHY: \
SWITCH_LOSS_FUNCTION_CASE(CauchyLoss); \
break; \
case BundleOptions::LossType::TRUNCATED_LE_ZACH: \
SWITCH_LOSS_FUNCTION_CASE(TruncatedLossLeZach); \
break;
///////////////////////////////////////////////////////////////////////////////////////////////////////
// Iteration callbacks (called after each LM iteration)
// Callback which prints debug info from the iterations
void print_iteration(const BundleStats &stats) {
if (stats.iterations == 0) {
std::cout << "initial_cost=" << stats.initial_cost << "\n";
}
std::cout << "iter=" << stats.iterations << ", cost=" << stats.cost << ", step=" << stats.step_norm
<< ", grad=" << stats.grad_norm << ", lambda=" << stats.lambda << "\n";
}
template <typename LossFunction> IterationCallback setup_callback(const BundleOptions &opt, LossFunction &loss_fn) {
if (opt.verbose) {
return print_iteration;
} else {
return nullptr;
}
}
// For using the IRLS scheme proposed by Le and Zach 3DV2021, we have a callback
// for each iteration which updates the mu parameter
template <> IterationCallback setup_callback(const BundleOptions &opt, TruncatedLossLeZach &loss_fn) {
if (opt.verbose) {
return [&loss_fn](const BundleStats &stats) {
print_iteration(stats);
loss_fn.mu *= TruncatedLossLeZach::alpha;
};
} else {
return [&loss_fn](const BundleStats &stats) { loss_fn.mu *= TruncatedLossLeZach::alpha; };
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////
// Absolute pose with points (PnP)
// Interface for calibrated camera
BundleStats bundle_adjust(const std::vector<Point2D> &x, const std::vector<Point3D> &X, CameraPose *pose,
const BundleOptions &opt, const std::vector<double> &weights) {
poselib::Camera camera;
camera.model_id = NullCameraModel::model_id;
return bundle_adjust(x, X, camera, pose, opt);
}
template <typename WeightType, typename CameraModel, typename LossFunction>
BundleStats bundle_adjust(const std::vector<Point2D> &x, const std::vector<Point3D> &X, const Camera &camera,
CameraPose *pose, const BundleOptions &opt, const WeightType &weights) {
LossFunction loss_fn(opt.loss_scale);
IterationCallback callback = setup_callback(opt, loss_fn);
CameraJacobianAccumulator<CameraModel, LossFunction, WeightType> accum(x, X, camera, loss_fn, weights);
return lm_impl<decltype(accum)>(accum, pose, opt, callback);
}
template <typename WeightType, typename CameraModel>
BundleStats bundle_adjust(const std::vector<Point2D> &x, const std::vector<Point3D> &X, const Camera &camera,
CameraPose *pose, const BundleOptions &opt, const WeightType &weights) {
switch (opt.loss_type) {
#define SWITCH_LOSS_FUNCTION_CASE(LossFunction) \
return bundle_adjust<WeightType, CameraModel, LossFunction>(x, X, camera, pose, opt, weights);
SWITCH_LOSS_FUNCTIONS
default:
return BundleStats();
}
#undef SWITCH_LOSS_FUNCTION_CASE
}
template <typename WeightType>
BundleStats bundle_adjust(const std::vector<Point2D> &x, const std::vector<Point3D> &X, const Camera &camera,
CameraPose *pose, const BundleOptions &opt, const WeightType &weights) {
switch (camera.model_id) {
#define SWITCH_CAMERA_MODEL_CASE(Model) \
case Model::model_id: { \
return bundle_adjust<WeightType, Model>(x, X, camera, pose, opt, weights); \
}
SWITCH_CAMERA_MODELS
#undef SWITCH_CAMERA_MODEL_CASE
default:
return BundleStats();
}
}
// Entry point for PnP refinement
BundleStats bundle_adjust(const std::vector<Point2D> &x, const std::vector<Point3D> &X, const Camera &camera,
CameraPose *pose, const BundleOptions &opt, const std::vector<double> &weights) {
if (weights.size() == x.size()) {
return bundle_adjust<std::vector<double>>(x, X, camera, pose, opt, weights);
} else {
return bundle_adjust<UniformWeightVector>(x, X, camera, pose, opt, UniformWeightVector());
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////
// Absolute pose with points and lines (PnPL)
// Note that we currently do not support different camera models here
// TODO: decide how to handle lines for non-linear camera models...
template <typename PointWeightType, typename LineWeightType, typename PointLossFunction, typename LineLossFunction>
BundleStats bundle_adjust(const std::vector<Point2D> &points2D, const std::vector<Point3D> &points3D,
const std::vector<Line2D> &lines2D, const std::vector<Line3D> &lines3D, CameraPose *pose,
const BundleOptions &opt, const BundleOptions &opt_line, const PointWeightType &weights_pts,
const LineWeightType &weights_lines) {
PointLossFunction pt_loss_fn(opt.loss_scale);
LineLossFunction line_loss_fn(opt_line.loss_scale);
IterationCallback callback = setup_callback(opt, pt_loss_fn);
PointLineJacobianAccumulator<PointLossFunction, LineLossFunction, PointWeightType, LineWeightType> accum(
points2D, points3D, lines2D, lines3D, pt_loss_fn, line_loss_fn, weights_pts, weights_lines);
return lm_impl<decltype(accum)>(accum, pose, opt, callback);
}
template <typename PointWeightType, typename LineWeightType, typename PointLossFunction>
BundleStats bundle_adjust(const std::vector<Point2D> &points2D, const std::vector<Point3D> &points3D,
const std::vector<Line2D> &lines2D, const std::vector<Line3D> &lines3D, CameraPose *pose,
const BundleOptions &opt, const BundleOptions &opt_line, const PointWeightType &weights_pts,
const LineWeightType &weights_lines) {
switch (opt_line.loss_type) {
#define SWITCH_LOSS_FUNCTION_CASE(LossFunction) \
return bundle_adjust<PointWeightType, LineWeightType, PointLossFunction, LossFunction>( \
points2D, points3D, lines2D, lines3D, pose, opt, opt_line, weights_pts, weights_lines);
SWITCH_LOSS_FUNCTIONS
default:
return BundleStats();
}
#undef SWITCH_LOSS_FUNCTION_CASE
}
template <typename PointWeightType, typename LineWeightType>
BundleStats bundle_adjust(const std::vector<Point2D> &points2D, const std::vector<Point3D> &points3D,
const std::vector<Line2D> &lines2D, const std::vector<Line3D> &lines3D, CameraPose *pose,
const BundleOptions &opt, const BundleOptions &opt_line, const PointWeightType &weights_pts,
const LineWeightType &weights_lines) {
switch (opt.loss_type) {
#define SWITCH_LOSS_FUNCTION_CASE(LossFunction) \
return bundle_adjust<PointWeightType, LineWeightType, LossFunction>(points2D, points3D, lines2D, lines3D, pose, \
opt, opt_line, weights_pts, weights_lines);
SWITCH_LOSS_FUNCTIONS
default:
return BundleStats();
}
#undef SWITCH_LOSS_FUNCTION_CASE
}
// Entry point for PnPL refinement
BundleStats bundle_adjust(const std::vector<Point2D> &points2D, const std::vector<Point3D> &points3D,
const std::vector<Line2D> &lines2D, const std::vector<Line3D> &lines3D, CameraPose *pose,
const BundleOptions &opt, const BundleOptions &opt_line,
const std::vector<double> &weights_pts, const std::vector<double> &weights_lines) {
bool have_pts_weights = weights_pts.size() == points2D.size();
bool have_line_weights = weights_lines.size() == lines2D.size();
if (have_pts_weights && have_line_weights) {
return bundle_adjust<std::vector<double>, std::vector<double>>(points2D, points3D, lines2D, lines3D, pose, opt,
opt_line, weights_pts, weights_lines);
} else if (have_pts_weights && !have_line_weights) {
return bundle_adjust<std::vector<double>, UniformWeightVector>(points2D, points3D, lines2D, lines3D, pose, opt,
opt_line, weights_pts, UniformWeightVector());
} else if (!have_pts_weights && have_line_weights) {
return bundle_adjust<UniformWeightVector, std::vector<double>>(points2D, points3D, lines2D, lines3D, pose, opt,
opt_line, UniformWeightVector(), weights_lines);
} else {
return bundle_adjust<UniformWeightVector, UniformWeightVector>(
points2D, points3D, lines2D, lines3D, pose, opt, opt_line, UniformWeightVector(), UniformWeightVector());
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////
// Generalized absolute pose with points (GPnP)
// Interface for calibrated camera
BundleStats generalized_bundle_adjust(const std::vector<std::vector<Point2D>> &x,
const std::vector<std::vector<Point3D>> &X,
const std::vector<CameraPose> &camera_ext, CameraPose *pose,
const BundleOptions &opt, const std::vector<std::vector<double>> &weights) {
std::vector<Camera> dummy_cameras;
dummy_cameras.resize(x.size());
for (size_t k = 0; k < x.size(); ++k) {
dummy_cameras[k].model_id = -1;
}
return generalized_bundle_adjust(x, X, camera_ext, dummy_cameras, pose, opt, weights);
}
template <typename WeightType, typename LossFunction>
BundleStats generalized_bundle_adjust(const std::vector<std::vector<Point2D>> &x,
const std::vector<std::vector<Point3D>> &X,
const std::vector<CameraPose> &camera_ext, const std::vector<Camera> &cameras,
CameraPose *pose, const BundleOptions &opt, const WeightType &weights) {
LossFunction loss_fn(opt.loss_scale);
IterationCallback callback = setup_callback(opt, loss_fn);
GeneralizedCameraJacobianAccumulator<LossFunction, WeightType> accum(x, X, camera_ext, cameras, loss_fn, weights);
return lm_impl<decltype(accum)>(accum, pose, opt, callback);
}
template <typename WeightType>
BundleStats generalized_bundle_adjust(const std::vector<std::vector<Point2D>> &x,
const std::vector<std::vector<Point3D>> &X,
const std::vector<CameraPose> &camera_ext, const std::vector<Camera> &cameras,
CameraPose *pose, const BundleOptions &opt, const WeightType &weights) {
switch (opt.loss_type) {
#define SWITCH_LOSS_FUNCTION_CASE(LossFunction) \
return generalized_bundle_adjust<WeightType, LossFunction>(x, X, camera_ext, cameras, pose, opt, weights);
SWITCH_LOSS_FUNCTIONS
default:
return BundleStats();
}
#undef SWITCH_LOSS_FUNCTION_CASE
}
// Entry point for GPnP refinement
BundleStats generalized_bundle_adjust(const std::vector<std::vector<Point2D>> &x,
const std::vector<std::vector<Point3D>> &X,
const std::vector<CameraPose> &camera_ext, const std::vector<Camera> &cameras,
CameraPose *pose, const BundleOptions &opt,
const std::vector<std::vector<double>> &weights) {
if (weights.size() == x.size()) {
return generalized_bundle_adjust<std::vector<std::vector<double>>>(x, X, camera_ext, cameras, pose, opt,
weights);
} else {
return generalized_bundle_adjust<UniformWeightVectors>(x, X, camera_ext, cameras, pose, opt,
UniformWeightVectors());
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////
// Relative pose (essential matrix) refinement
template <typename WeightType, typename LossFunction>
BundleStats refine_relpose(const std::vector<Point2D> &x1, const std::vector<Point2D> &x2, CameraPose *pose,
const BundleOptions &opt, const WeightType &weights) {
LossFunction loss_fn(opt.loss_scale);
IterationCallback callback = setup_callback(opt, loss_fn);
RelativePoseJacobianAccumulator<LossFunction, WeightType> accum(x1, x2, loss_fn, weights);
return lm_impl<decltype(accum)>(accum, pose, opt, callback);
}
template <typename WeightType>
BundleStats refine_relpose(const std::vector<Point2D> &x1, const std::vector<Point2D> &x2, CameraPose *pose,
const BundleOptions &opt, const WeightType &weights) {
switch (opt.loss_type) {
#define SWITCH_LOSS_FUNCTION_CASE(LossFunction) \
return refine_relpose<WeightType, LossFunction>(x1, x2, pose, opt, weights);
SWITCH_LOSS_FUNCTIONS
default:
return BundleStats();
}
#undef SWITCH_LOSS_FUNCTION_CASE
}
// Entry point for essential matrix refinement
BundleStats refine_relpose(const std::vector<Point2D> &x1, const std::vector<Point2D> &x2, CameraPose *pose,
const BundleOptions &opt, const std::vector<double> &weights) {
if (weights.size() == x1.size()) {
return refine_relpose<std::vector<double>>(x1, x2, pose, opt, weights);
} else {
return refine_relpose<UniformWeightVector>(x1, x2, pose, opt, UniformWeightVector());
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////
// Relative pose (essential matrix) refinement
template <typename WeightType, typename LossFunction>
BundleStats refine_shared_focal_relpose(const std::vector<Point2D> &x1, const std::vector<Point2D> &x2,
ImagePair *image_pair, const BundleOptions &opt, const WeightType &weights) {
LossFunction loss_fn(opt.loss_scale);
IterationCallback callback = setup_callback(opt, loss_fn);
SharedFocalRelativePoseJacobianAccumulator<LossFunction, WeightType> accum(x1, x2, loss_fn, weights);
return lm_impl<decltype(accum)>(accum, image_pair, opt, callback);
}
template <typename WeightType>
BundleStats refine_shared_focal_relpose(const std::vector<Point2D> &x1, const std::vector<Point2D> &x2,
ImagePair *image_pair, const BundleOptions &opt, const WeightType &weights) {
switch (opt.loss_type) {
#define SWITCH_LOSS_FUNCTION_CASE(LossFunction) \
return refine_shared_focal_relpose<WeightType, LossFunction>(x1, x2, image_pair, opt, weights);
SWITCH_LOSS_FUNCTIONS
default:
return BundleStats();
}
#undef SWITCH_LOSS_FUNCTION_CASE
}
// Entry point for essential matrix refinement
BundleStats refine_shared_focal_relpose(const std::vector<Point2D> &x1, const std::vector<Point2D> &x2,
ImagePair *image_pair, const BundleOptions &opt,
const std::vector<double> &weights) {
if (weights.size() == x1.size()) {
return refine_shared_focal_relpose<std::vector<double>>(x1, x2, image_pair, opt, weights);
} else {
return refine_shared_focal_relpose<UniformWeightVector>(x1, x2, image_pair, opt, UniformWeightVector());
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////
// Uncalibrated relative pose (fundamental matrix) refinement
template <typename WeightType, typename LossFunction>
BundleStats refine_fundamental(const std::vector<Point2D> &x1, const std::vector<Point2D> &x2, Eigen::Matrix3d *F,
const BundleOptions &opt, const WeightType &weights) {
// We optimize over the SVD-based factorization from Bartoli and Sturm
FactorizedFundamentalMatrix factorized_fund_mat(*F);
LossFunction loss_fn(opt.loss_scale);
IterationCallback callback = setup_callback(opt, loss_fn);
FundamentalJacobianAccumulator<LossFunction, WeightType> accum(x1, x2, loss_fn, weights);
BundleStats stats = lm_impl<decltype(accum)>(accum, &factorized_fund_mat, opt, callback);
*F = factorized_fund_mat.F();
return stats;
}
template <typename WeightType>
BundleStats refine_fundamental(const std::vector<Point2D> &x1, const std::vector<Point2D> &x2, Eigen::Matrix3d *F,
const BundleOptions &opt, const WeightType &weights) {
switch (opt.loss_type) {
#define SWITCH_LOSS_FUNCTION_CASE(LossFunction) \
return refine_fundamental<WeightType, LossFunction>(x1, x2, F, opt, weights);
SWITCH_LOSS_FUNCTIONS
default:
return BundleStats();
}
#undef SWITCH_LOSS_FUNCTION_CASE
}
// Entry point for fundamental matrix refinement
BundleStats refine_fundamental(const std::vector<Point2D> &x1, const std::vector<Point2D> &x2, Eigen::Matrix3d *F,
const BundleOptions &opt, const std::vector<double> &weights) {
if (weights.size() == x1.size()) {
return refine_fundamental<std::vector<double>>(x1, x2, F, opt, weights);
} else {
return refine_fundamental<UniformWeightVector>(x1, x2, F, opt, UniformWeightVector());
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////
// Homography matrix refinement
template <typename WeightType, typename LossFunction>
BundleStats refine_homography(const std::vector<Point2D> &x1, const std::vector<Point2D> &x2, Eigen::Matrix3d *H,
const BundleOptions &opt, const WeightType &weights) {
LossFunction loss_fn(opt.loss_scale);
IterationCallback callback = setup_callback(opt, loss_fn);
HomographyJacobianAccumulator<LossFunction, WeightType> accum(x1, x2, loss_fn, weights);
return lm_impl<decltype(accum)>(accum, H, opt, callback);
}
template <typename WeightType>
BundleStats refine_homography(const std::vector<Point2D> &x1, const std::vector<Point2D> &x2, Eigen::Matrix3d *H,
const BundleOptions &opt, const WeightType &weights) {
switch (opt.loss_type) {
#define SWITCH_LOSS_FUNCTION_CASE(LossFunction) \
return refine_homography<WeightType, LossFunction>(x1, x2, H, opt, weights);
SWITCH_LOSS_FUNCTIONS
default:
return BundleStats();
}
#undef SWITCH_LOSS_FUNCTION_CASE
}
// Entry point for fundamental matrix refinement
BundleStats refine_homography(const std::vector<Point2D> &x1, const std::vector<Point2D> &x2, Eigen::Matrix3d *H,
const BundleOptions &opt, const std::vector<double> &weights) {
if (weights.size() == x1.size()) {
return refine_homography<std::vector<double>>(x1, x2, H, opt, weights);
} else {
return refine_homography<UniformWeightVector>(x1, x2, H, opt, UniformWeightVector());
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////
// Generalized relative pose refinement
template <typename WeightType, typename LossFunction>
BundleStats refine_generalized_relpose(const std::vector<PairwiseMatches> &matches,
const std::vector<CameraPose> &camera1_ext,
const std::vector<CameraPose> &camera2_ext, CameraPose *pose,
const BundleOptions &opt, const WeightType &weights) {
LossFunction loss_fn(opt.loss_scale);
IterationCallback callback = setup_callback(opt, loss_fn);
GeneralizedRelativePoseJacobianAccumulator<LossFunction, WeightType> accum(matches, camera1_ext, camera2_ext,
loss_fn, weights);
return lm_impl<decltype(accum)>(accum, pose, opt, callback);
}
template <typename WeightType>
BundleStats refine_generalized_relpose(const std::vector<PairwiseMatches> &matches,
const std::vector<CameraPose> &camera1_ext,
const std::vector<CameraPose> &camera2_ext, CameraPose *pose,
const BundleOptions &opt, const WeightType &weights) {
switch (opt.loss_type) {
#define SWITCH_LOSS_FUNCTION_CASE(LossFunction) \
return refine_generalized_relpose<WeightType, LossFunction>(matches, camera1_ext, camera2_ext, pose, opt, weights);
SWITCH_LOSS_FUNCTIONS
default:
return BundleStats();
}
#undef SWITCH_LOSS_FUNCTION_CASE
}
// Entry point for generalized relpose refinement
BundleStats refine_generalized_relpose(const std::vector<PairwiseMatches> &matches,
const std::vector<CameraPose> &camera1_ext,
const std::vector<CameraPose> &camera2_ext, CameraPose *pose,
const BundleOptions &opt, const std::vector<std::vector<double>> &weights) {
if (weights.size() == matches.size()) {
return refine_generalized_relpose<std::vector<std::vector<double>>>(matches, camera1_ext, camera2_ext, pose,
opt, weights);
} else {
return refine_generalized_relpose<UniformWeightVectors>(matches, camera1_ext, camera2_ext, pose, opt,
UniformWeightVectors());
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////
// Hybrid pose refinement (i.e. both 2D-3D and 2D-2D point correspondences)
template <typename AbsWeightType, typename RelWeightType, typename LossFunction>
BundleStats refine_hybrid_pose(const std::vector<Point2D> &x, const std::vector<Point3D> &X,
const std::vector<PairwiseMatches> &matches_2D_2D,
const std::vector<CameraPose> &map_ext, CameraPose *pose, const BundleOptions &opt,
double loss_scale_epipolar, const AbsWeightType &weights_abs,
const RelWeightType &weights_rel) {
LossFunction loss_fn(opt.loss_scale);
LossFunction loss_fn_epipolar(loss_scale_epipolar);
// TODO: refactor such that the callback can handle multiple loss-functions
// currently this only affects TruncatedLossLeZach
IterationCallback callback = setup_callback(opt, loss_fn);
HybridPoseJacobianAccumulator<LossFunction, AbsWeightType, RelWeightType> accum(
x, X, matches_2D_2D, map_ext, loss_fn, loss_fn_epipolar, weights_abs, weights_rel);
return lm_impl<decltype(accum)>(accum, pose, opt, callback);
}
template <typename AbsWeightType, typename RelWeightType>
BundleStats refine_hybrid_pose(const std::vector<Point2D> &x, const std::vector<Point3D> &X,
const std::vector<PairwiseMatches> &matches_2D_2D,
const std::vector<CameraPose> &map_ext, CameraPose *pose, const BundleOptions &opt,
double loss_scale_epipolar, const AbsWeightType &weights_abs,
const RelWeightType &weights_rel) {
switch (opt.loss_type) {
#define SWITCH_LOSS_FUNCTION_CASE(LossFunction) \
return refine_hybrid_pose<AbsWeightType, RelWeightType, LossFunction>( \
x, X, matches_2D_2D, map_ext, pose, opt, loss_scale_epipolar, weights_abs, weights_rel);
SWITCH_LOSS_FUNCTIONS
default:
return BundleStats();
}
#undef SWITCH_LOSS_FUNCTION_CASE
}
// Entry point for hybrid pose refinement
BundleStats refine_hybrid_pose(const std::vector<Point2D> &x, const std::vector<Point3D> &X,
const std::vector<PairwiseMatches> &matches_2D_2D,
const std::vector<CameraPose> &map_ext, CameraPose *pose, const BundleOptions &opt,
double loss_scale_epipolar, const std::vector<double> &weights_abs,
const std::vector<std::vector<double>> &weights_rel) {
bool have_abs_weights = weights_abs.size() == x.size();
bool have_rel_weights = weights_rel.size() == matches_2D_2D.size();
if (have_abs_weights && have_rel_weights) {
return refine_hybrid_pose<std::vector<double>, std::vector<std::vector<double>>>(
x, X, matches_2D_2D, map_ext, pose, opt, loss_scale_epipolar, weights_abs, weights_rel);
} else if (have_abs_weights && !have_rel_weights) {
return refine_hybrid_pose<std::vector<double>, UniformWeightVectors>(
x, X, matches_2D_2D, map_ext, pose, opt, loss_scale_epipolar, weights_abs, UniformWeightVectors());
} else if (!have_abs_weights && have_rel_weights) {
return refine_hybrid_pose<UniformWeightVector, std::vector<std::vector<double>>>(
x, X, matches_2D_2D, map_ext, pose, opt, loss_scale_epipolar, UniformWeightVector(), weights_rel);
} else {
return refine_hybrid_pose<UniformWeightVector, UniformWeightVectors>(x, X, matches_2D_2D, map_ext, pose, opt,
loss_scale_epipolar, UniformWeightVector(),
UniformWeightVectors());
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////
// 1D-radial absolute pose refinement (1D-radial PnP)
template <typename WeightType, typename LossFunction>
BundleStats bundle_adjust_1D_radial(const std::vector<Point2D> &x, const std::vector<Point3D> &X, CameraPose *pose,
const BundleOptions &opt, const WeightType &weights) {
LossFunction loss_fn(opt.loss_scale);
IterationCallback callback = setup_callback(opt, loss_fn);
Radial1DJacobianAccumulator<LossFunction, WeightType> accum(x, X, loss_fn, weights);
return lm_impl<decltype(accum)>(accum, pose, opt, callback);
}
template <typename WeightType>
BundleStats bundle_adjust_1D_radial(const std::vector<Point2D> &x, const std::vector<Point3D> &X, CameraPose *pose,
const BundleOptions &opt, const WeightType &weights) {
switch (opt.loss_type) {
#define SWITCH_LOSS_FUNCTION_CASE(LossFunction) \
return bundle_adjust_1D_radial<WeightType, LossFunction>(x, X, pose, opt, weights);
SWITCH_LOSS_FUNCTIONS
default:
return BundleStats();
}
#undef SWITCH_LOSS_FUNCTION_CASE
}
// Entry point for 1D radial absolute pose refinement (Assumes that the image points are centered)
BundleStats bundle_adjust_1D_radial(const std::vector<Point2D> &x, const std::vector<Point3D> &X, CameraPose *pose,
const BundleOptions &opt, const std::vector<double> &weights) {
if (weights.size() == x.size()) {
return bundle_adjust_1D_radial<std::vector<double>>(x, X, pose, opt, weights);
} else {
return bundle_adjust_1D_radial<UniformWeightVector>(x, X, pose, opt, UniformWeightVector());
}
}
#undef SWITCH_LOSS_FUNCTIONS
} // namespace poselib
|
