-
Notifications
You must be signed in to change notification settings - Fork 40
/
CIWT.cpp
736 lines (607 loc) · 36.6 KB
/
CIWT.cpp
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
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
/*
Copyright 2017. All rights reserved.
Computer Vision Group, Visual Computing Institute
RWTH Aachen University, Germany
This file is part of the rwth_mot framework.
Authors: Aljosa Osep (osep -at- vision.rwth-aachen.de)
rwth_mot framework is free software; you can redistribute it and/or modify it under the
terms of the GNU General Public License as published by the Free Software
Foundation; either version 3 of the License, or any later version.
rwth_mot framework is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with
rwth_mot framework; if not, write to the Free Software Foundation, Inc., 51 Franklin
Street, Fifth Floor, Boston, MA 02110-1301, USA
*/
// C
#include <ctime>
// std
#include <iostream>
#include <memory>
#include <algorithm>
#include <list>
#include <chrono>
// opencv
#include <opencv2/core/core.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/imgproc/imgproc.hpp>
// pcl
#include <pcl/common/common_headers.h>
#include <pcl/common/transforms.h>
#include <pcl/io/pcd_io.h>
#include <pcl/io/ply_io.h>
// boost
#include <boost/archive/binary_iarchive.hpp>
// scene segmentation
#include <scene_segmentation/scene_segmentation.h>
#include <scene_segmentation/utils_segmentation.h>
#include <scene_segmentation/multi_scale_quickshift.h>
#include <scene_segmentation/parameters_gop3D.h>
// tracking
#include <tracking/visualization.h>
#include <tracking/utils_tracking.h>
#include <src/sun_utils/detection.h>
#include <tracking/category_filter.h>
// utils
#include "utils_io.h"
#include "utils_visualization.h"
#include "utils_pointcloud.h"
#include "ground_model.h"
#include "utils_observations.h"
#include "datasets_dirty_utils.h"
#include "utils_bounding_box.h"
#include "utils_common.h"
#include "utils_filtering.h"
// CIWT
#include "CIWT/parameters_CIWT.h"
#include "CIWT/CIWT_tracker.h"
#include "CIWT/observation_fusion.h"
#include "CIWT/potential_functions.h"
#define MAX_PATH_LEN 500
// For convenience.
namespace po = boost::program_options;
typedef pcl::PointCloud<pcl::PointXYZRGBA> PointCloudRGBA;
namespace CIWTApp {
// We need those for the 3d visualizer thread.
bool visualization_3d_update_flag;
boost::mutex visualization_3d_update_mutex;
PointCloudRGBA::Ptr visualization_3d_point_cloud;
GOT::tracking::HypothesesVector visualization_3d_tracking_hypotheses;
GOT::tracking::HypothesesVector visualization_3d_tracking_terminated_hypotheses;
std::vector<GOT::tracking::Observation> visualization_observations;
std::vector<GOT::segmentation::ObjectProposal> visualization_3d_proposals;
SUN::utils::Camera visualization_3d_left_camera;
GOT::tracking::Visualizer tracking_visualizer;
// Paths
std::string output_dir;
std::string proposals_path;
std::string tracking_mode_str;
std::string viewer_3D_output_path;
std::string config_parameters_file;
std::string sequence_name;
std::string dataset_name;
// Application data.
bool show_visualization_2d;
bool show_visualization_3d;
int start_frame;
int end_frame;
int debug_level;
bool run_tracker;
Eigen::Matrix4d egomotion = Eigen::Matrix4d::Identity();
void VisualizeScene3D() {
// Set up the visualizer.
pcl::visualization::PCLVisualizer viewer("3D Scene Viewer");
pcl::visualization::PointCloudColorHandlerRGBField<pcl::PointXYZRGBA> rgbHandler(visualization_3d_point_cloud);
viewer.setPointCloudRenderingProperties (pcl::visualization::PCL_VISUALIZER_POINT_SIZE, 3, "sample cloud");
viewer.setBackgroundColor (0.0, 0.0, 0.0);
viewer.setCameraPosition(-0.643694,-1.45647,-4.11917, //camera optical center position
-0.0132233,-0.999894,0.0060805, //Look-at-position
0, -1, 0); //camera up vector
viewer.setCameraFieldOfView(0.8575);
viewer.setCameraClipDistances(0.0001, 900);
viewer.setSize(2000, 800);
// Visualizer main loop.
while (!viewer.wasStopped ()) {
viewer.spinOnce(100);
// Get lock on the boolean update and check if cloud was updated.
boost::mutex::scoped_lock updateLock(visualization_3d_update_mutex);
if (visualization_3d_update_flag) {
viewer.removeAllShapes();
viewer.removeAllPointClouds();
// Show point-cloud.
if(!viewer.updatePointCloud(visualization_3d_point_cloud, rgbHandler, "sample cloud")) {
viewer.addPointCloud(visualization_3d_point_cloud, rgbHandler, "sample cloud");
viewer.setPointCloudRenderingProperties (pcl::visualization::PCL_VISUALIZER_POINT_SIZE, 4, "sample cloud");
viewer.setPointCloudRenderingProperties (pcl::visualization::PCL_VISUALIZER_OPACITY, 0.7, "sample cloud");
}
// Settings
bool draw_observations = false;
bool draw_tracked_objects = true;
bool draw_terminated_hypos = false;
bool mark_only_proposal_region = false;
bool draw_proposals = false;
if (draw_proposals) {
// Render 3D bounding boxes.
for (int i=0; i<visualization_3d_proposals.size(); i++) {
// 3D bounding box: [centroid_x centroid_y centroid_z width height length q.w q.x q.y q.z]
auto &proposal = visualization_3d_proposals.at(i);
auto bbox_3d = proposal.bounding_box_3d();
std::string id = "bb_3d_" + std::to_string(i);
SUN::utils::visualization::RenderBoundingBox3D(viewer, bbox_3d, 0.0, 0.0, 1.0, id);
}
}
if (draw_observations) {
// Draw observations
for (int i=0; i<visualization_observations.size(); i++) {
const auto &obs = visualization_observations.at(i);
uint8_t r, g, b;
SUN::utils::visualization::GenerateColor(i,r,g,b);
double r_ = static_cast<double> (r) / 255.0;
double g_ = static_cast<double> (g) / 255.0;
double b_ = static_cast<double> (b) / 255.0;
PointCloudRGBA::Ptr seg_cloud(new PointCloudRGBA);
if (mark_only_proposal_region) {
// Draw only 'associated' detection part.
pcl::PointIndices inds;
inds.indices = obs.pointcloud_indices();
pcl::copyPointCloud(*visualization_3d_point_cloud, inds, *seg_cloud);
}
else {
// For ICRA'17 paper visualization: full-projection of bounding-box to point-cloud.
Eigen::Vector4d det_bbox_2d = obs.bounding_box_2d();
for (int i=det_bbox_2d[0]; i<det_bbox_2d[2]+det_bbox_2d[0]; i++) {
for (int j=det_bbox_2d[1]; j<det_bbox_2d[3]+det_bbox_2d[1]; j++) {
pcl::PointXYZRGBA p = visualization_3d_point_cloud->at(i,j);
if (!std::isnan(p.x)) {
p.r = r; p.b = b; p.g = g;
seg_cloud->push_back(p);
}
}
}
seg_cloud->height = 1;
seg_cloud->width = seg_cloud->size();
}
pcl::visualization::PointCloudColorHandlerRGBField<pcl::PointXYZRGBA> seg_handler(seg_cloud);
auto seg_cloud_id = "seg_"+std::to_string(i);
if(!viewer.updatePointCloud(seg_cloud, seg_handler, seg_cloud_id)) {
viewer.addPointCloud(seg_cloud, seg_handler, seg_cloud_id);
viewer.setPointCloudRenderingProperties (pcl::visualization::PCL_VISUALIZER_POINT_SIZE, 4, seg_cloud_id);
viewer.setPointCloudRenderingProperties (pcl::visualization::PCL_VISUALIZER_OPACITY, 0.5, seg_cloud_id);
viewer.setPointCloudRenderingProperties (pcl::visualization::PCL_VISUALIZER_COLOR, r_, g_, b_, seg_cloud_id);
}
}
}
if (draw_tracked_objects) {
// Draw active hypotheses.
for (int i=0; i<visualization_3d_tracking_hypotheses.size(); i++) {
const auto &hypo = visualization_3d_tracking_hypotheses.at(i);
double r,g,b;
tracking_visualizer.GetColor(hypo.id(), b,g,r);
if (hypo.bounding_boxes_3d().size()>0) {
std::string id = "hypo_bbox_3d_" + std::to_string(i);
tracking_visualizer.RenderHypo3D(viewer, hypo, CIWTApp::visualization_3d_left_camera, 0);
}
tracking_visualizer.RenderTrajectory(hypo, CIWTApp::visualization_3d_left_camera, ("hypo_traj_3d_" + std::to_string(i)), r,g,b, viewer);
}
// Draw terminated hypotheses.
if (draw_terminated_hypos) {
for (int i=0; i<visualization_3d_tracking_terminated_hypotheses.size(); i++) {
const auto &hypo = visualization_3d_tracking_terminated_hypotheses.at(i);
if (hypo.bounding_boxes_3d().size() > 0) {
std::string id = "hypo_term_bbox_3d_" + std::to_string(i);
SUN::utils::visualization::RenderBoundingBox3D(viewer, hypo.bounding_boxes_3d().back(), 0.0, 0.0, 1.0, id, 0);
}
tracking_visualizer.RenderTrajectory(hypo, CIWTApp::visualization_3d_left_camera, ("hypo_term_traj_3d_" + std::to_string(i)), 0.0, 0.0, 1.0, viewer);
}
}
}
if (debug_level>=3)
viewer.saveScreenshot(viewer_3D_output_path);
visualization_3d_update_flag = false;
viewer.setCameraClipDistances(0.0001, 900);
}
updateLock.unlock();
}
}
bool ParseCommandArguments(const int argc, const char **argv, po::variables_map &config_variables_map) {
po::options_description cmdline_options;
try {
std::string config_file;
// The very basic options, can be only specified via cmd args
po::options_description generic_options("Command line options:");
generic_options.add_options()
("help", "Produce help message")
("config", po::value<std::string>(&config_file)->required(), "Config file path.")
("config_parameters", po::value<std::string>(&config_parameters_file), "Config file path (parameters only!).")
("show_visualization_3d", po::bool_switch(&show_visualization_3d)->default_value(false), "Show 3D visualization.")
("show_visualization_2d", po::bool_switch(&show_visualization_2d)->default_value(false), "Show 2D visualization.")
("run_tracker", po::value<bool>(&run_tracker)->default_value(true), "Should run tracker, or should not?")
("debug_level", po::value<int>(&debug_level)->default_value(0), "Debug level")
;
// General app options (can be spec. in config or via cmd args)
po::options_description config_options("Config options");
config_options.add_options()
// Paths
("left_image_path", po::value<std::string>(), "Image (left) path")
("right_image_path", po::value<std::string>(), "Image (right) path")
("left_disparity_path", po::value<std::string>(), "Disparity (left) path")
("calib_path", po::value<std::string>(), "Camera calibration path (currently supported: kitti)")
("object_proposal_path", po::value<std::string>(&proposals_path)->default_value(""), "Proposals path")
("ground_plane_path", po::value<std::string>(), "Ground-plane params.")
("detections_path", po::value<std::string>()->default_value(""), "Resources path.")
("flow_map_path", po::value<std::string>(), "Path to flow dir.")
("output_dir", po::value<std::string>(&output_dir), "Output path")
("start_frame", po::value<int>(&start_frame)->default_value(0), "Starting frame")
("end_frame", po::value<int>(&end_frame)->default_value(10000), "Last frame")
("tracking_mode", po::value<std::string>(&tracking_mode_str)->default_value("detection"), "Tracking mode: detection, detection_shape, detection_3DOP.")
("sequence_name", po::value<std::string>(&sequence_name)->default_value("default_sequence_name"), "Name of the sequence")
("dataset_name", po::value<std::string>(&dataset_name)->default_value("kitti"), "Name of the sequence")
;
// Add parameter options (def. in parameters_CIWT.h)
po::options_description parameter_options("Parameters:");
InitParameters(parameter_options);
// Parameters for the 3D proposals
po::options_description parameters_proposals("Parameters-proposals:");
GOP3D::InitParameters(parameters_proposals);
cmdline_options.add(generic_options);
cmdline_options.add(config_options);
cmdline_options.add(parameter_options);
cmdline_options.add(parameters_proposals);
store(po::command_line_parser(argc, argv).options(cmdline_options).run(), config_variables_map);
if (config_variables_map.count("help")) {
std::cout << cmdline_options << endl;
return false;
}
notify(config_variables_map);
// "generic" config
if (config_variables_map.count("config")) {
std::ifstream ifs(config_file.c_str());
if (!ifs.is_open()) {
std::cout << "Can not Open config file: " << config_file << "\n";
return false;
} else {
store(parse_config_file(ifs, cmdline_options), config_variables_map);
notify(config_variables_map);
}
}
if (config_variables_map.count("config_parameters")) {
// "parameter" config
std::ifstream ifs_param(config_parameters_file.c_str());
if (!ifs_param.is_open()) {
std::cout << "Can not Open parameter config file: " << config_parameters_file << "\n";
return false;
} else {
store(parse_config_file(ifs_param, cmdline_options), config_variables_map);
notify(config_variables_map);
}
}
}
catch(po::error& e) {
std::cerr << "ERROR: " << e.what() << std::endl << std::endl;
std::cerr << cmdline_options << std::endl;
return false;
}
return true;
}
bool RequestObjectProposals(int frame, po::variables_map &options_map,
std::function<std::vector<GOT::segmentation::ObjectProposal>(po::variables_map &)> proposal_gen_fnc,
std::vector<GOT::segmentation::ObjectProposal> &proposals_out,
bool save_if_not_avalible=true) {
// Check if proposals exist.
// Yes -> load.
// No -> run proposal generator.
if (options_map.count("object_proposal_path")) {
// There is path, see if files exist first, otherwise compute (and optionally store).
char proposal_path_buff[MAX_PATH_LEN];
snprintf(proposal_path_buff, MAX_PATH_LEN, options_map.at("object_proposal_path").as<std::string>().c_str(), frame);
auto success_loading_proposals = GOT::segmentation::utils::LoadObjectProposals(proposal_path_buff, proposals_out);
if (!success_loading_proposals) {
printf("Could not load proposals, computing (note: processing will slow-down!) ...\r\n");
proposals_out = proposal_gen_fnc(options_map);
if (save_if_not_avalible) {
std::cout << "Saving proposals to: " << proposal_path_buff << std::endl;
boost::filesystem::path prop_path(proposal_path_buff);
boost::filesystem::path prop_dir = prop_path.parent_path();
SUN::utils::IO::MakeDir(prop_dir.c_str());
GOT::segmentation::utils::SaveObjectProposals(proposal_path_buff, proposals_out);
}
}
return true;
}
return false;
}
}
/*
-------------
Debug Levels:
-------------
0 - Outputs basically nothing, except relevant error messages.
1 - Console output, logging.
2 - Quantitative evaluation.
3 - Most relevant visual results (per-frame, eg. segmentation, tracking results, ...).
4 - Point clouds (per-frame), less relevant visual results.
5 - Additional possibly relevant frame output (segmentation 3D data, integrated models, ...).
>=6 - All possible/necessary debug stuff. Should make everything really really really slow.
*/
int main(const int argc, const char** argv) {
std::cout << "Hello from CIWT!" << std::endl;
// -------------------------------------------------------------------------------
// +++ Command Args Parser +++
// -------------------------------------------------------------------------------
po::variables_map variables_map;
if (!CIWTApp::ParseCommandArguments(argc, argv, variables_map)) {
std::cout << "Failed at ParseCommandArguments ... " << std::endl;
return -1;
}
printf("Tracking mode: %s.\r\n", CIWTApp::tracking_mode_str.c_str());
printf("Run tracker: %s.\r\n", CIWTApp::run_tracker ? "YES" : "NO");
// -------------------------------------------------------------------------------
// +++ Globals +++
// -------------------------------------------------------------------------------
const int num_frames = CIWTApp::end_frame-CIWTApp::start_frame; // This many frames will be processed.
// Makes sure the relevant values are correct.
assert(num_frames>0);
assert(CIWTApp::debug_level>=0);
// Quantitative evaluation result storage
std::vector<SUN::utils::KITTI::TrackingLabel> tracker_result_labels;
// -------------------------------------------------------------------------------
// +++ Create output dirs +++
// -------------------------------------------------------------------------------
// You can add more output sub-dir's!
std::string output_dir_visual_results;
std::string output_dir_tracking_data;
if (CIWTApp::debug_level>=3) {
output_dir_visual_results = CIWTApp::output_dir + "/visual_results";
bool make_dir_success = SUN::utils::IO::MakeDir(output_dir_visual_results.c_str());
assert(make_dir_success);
output_dir_tracking_data = CIWTApp::output_dir + "/tracking_data";
make_dir_success = SUN::utils::IO::MakeDir(output_dir_tracking_data.c_str());
assert(make_dir_success);
}
// -------------------------------------------------------------------------------
// +++ Data handling +++
// -------------------------------------------------------------------------------
SUN::utils::dirty::DatasetAssitantDirty dataset_assistant(variables_map); // Data loading hacky module
PointCloudRGBA::Ptr left_point_cloud_preprocessed; // Preprocessed cloud
std::vector<GOT::segmentation::ObjectProposal> object_proposals_all; // 'Raw' object proposals set
std::shared_ptr<GOT::tracking::Resources> resource_manager(new GOT::tracking::Resources(variables_map["tracking_temporal_window_size"].as<int>()+1));
// -------------------------------------------------------------------------------
// +++ Visual odometry module +++
// -------------------------------------------------------------------------------
std::shared_ptr<libviso2::VisualOdometryStereo> vo_module = nullptr;
auto InitVO = [](std::shared_ptr<libviso2::VisualOdometryStereo> &vo, double f, double c_u, double c_v, double baseline) {
if (vo==nullptr) {
libviso2::VisualOdometryStereo::parameters param;
param.calib.f = f;
param.calib.cu = c_u;
param.calib.cv = c_v;
param.base = baseline;
vo.reset(new libviso2::VisualOdometryStereo(param));
}
};
// -------------------------------------------------------------------------------
// +++ Tracker +++
// -------------------------------------------------------------------------------
// Create the tracker object
std::unique_ptr<GOT::tracking::ciwt_tracker::CIWTTracker> multi_object_tracker(new GOT::tracking::ciwt_tracker::CIWTTracker(variables_map));
if (CIWTApp::debug_level>=3)
multi_object_tracker->set_verbose(true);
// -------------------------------------------------------------------------------
// +++ Visualization Threads +++
// -------------------------------------------------------------------------------
/// 3D visualizer thread
CIWTApp::visualization_3d_point_cloud.reset(new PointCloudRGBA);
std::unique_ptr<boost::thread> visualization_3d_thread;
if (CIWTApp::show_visualization_3d) {
visualization_3d_thread.reset(new boost::thread(CIWTApp::VisualizeScene3D));
}
/// 2D visualization threads
if (CIWTApp::show_visualization_2d) {
cv::namedWindow("tracking_2d_window");
cv::startWindowThread();
cv::namedWindow("observations_id_window");
cv::startWindowThread();
}
// -------------------------------------------------------------------------------
// +++ MAIN TRACKING LOOP +++
// -------------------------------------------------------------------------------
double total_processing_time = 0.0;
for (int current_frame=CIWTApp::start_frame; current_frame<=CIWTApp::end_frame; current_frame++) {
if(CIWTApp::debug_level>0){
std::cout << "\33[33;40;1m" << "-----------------------------------------------------------------------------" << "\33[0m" << std::endl;
std::cout << "\33[33;40;1m" <<" Processing frame " << current_frame << "\33[0m"<< std::endl;
std::cout << "\33[33;40;1m" << "-----------------------------------------------------------------------------" << "\33[0m" << std::endl;
}
// -------------------------------------------------------------------------------
// +++ Load data +++
// -------------------------------------------------------------------------------
if (!dataset_assistant.LoadData(current_frame, variables_map["dataset_name"].as<std::string>())) {
printf("Dataset assistant can't load data :'( Check your config. \r\n");
return -1;
}
auto &left_camera = dataset_assistant.left_camera_;
auto &right_camera = dataset_assistant.right_camera_;
const auto &left_image = dataset_assistant.left_image_;
const auto &planar_ground_model = left_camera.ground_model();
const auto &velocity_map = dataset_assistant.velocity_map_;
left_point_cloud_preprocessed.reset(new PointCloudRGBA);
pcl::copyPointCloud(*dataset_assistant.left_point_cloud_, *left_point_cloud_preprocessed);
std::vector<SUN::utils::Detection> detections_current_frame = dataset_assistant.object_detections_;
// -------------------------------------------------------------------------------
// +++ Run visual odometry module => estimate egomotion +++
// -------------------------------------------------------------------------------
// Estimate ego
InitVO(vo_module, left_camera.f_u(), left_camera.c_u(), left_camera.c_v(), dataset_assistant.stereo_baseline_); // Will be only initialized once, but need to do it within the loop
Eigen::Matrix4d ego_this_frame = GOT::tracking::utils::EstimateEgomotion(*vo_module, dataset_assistant.left_image_, dataset_assistant.right_image_);
// Accumulated transformation
CIWTApp::egomotion = CIWTApp::egomotion * ego_this_frame.inverse();
// Update left_camera, right_camera using estimated pose transform
left_camera.ApplyPoseTransform(CIWTApp::egomotion);
right_camera.ApplyPoseTransform(CIWTApp::egomotion);
// -------------------------------------------------------------------------------
// +++ 3D proposals +++
// -------------------------------------------------------------------------------
if(CIWTApp::debug_level>0) printf("->Processing object proposals ...\r\n");
bool success_loading_proposals = false;
if (CIWTApp::tracking_mode_str == "detection_shape") { // In detection-only mode, don't bother with proposals
success_loading_proposals = CIWTApp::RequestObjectProposals(current_frame, variables_map,
std::bind(GOT::segmentation::proposal_generation::ComputeSuppressedMultiScale3DProposals,
dataset_assistant.left_point_cloud_, dataset_assistant.left_camera_, dataset_assistant.right_camera_, std::placeholders::_1),
object_proposals_all);
assert(success_loading_proposals);
}
// Filter certain 'noise' proposals (reject small ones and flying-ones)
object_proposals_all = GOT::segmentation::utils::FilterProposals(object_proposals_all, left_camera, variables_map);
// Sort (or: Re-Sort) proposals by their score.
std::sort(object_proposals_all.begin(), object_proposals_all.end(),
[](const GOT::segmentation::ObjectProposal &i, const GOT::segmentation::ObjectProposal &j) { return i.score() > j.score(); });
if(CIWTApp::debug_level>0) printf("->Got %d proposals.\r\n", static_cast<int>(object_proposals_all.size()));
// Start per-frame timing analysis here!
std::chrono::steady_clock::time_point time_begin = std::chrono::steady_clock::now();
// -------------------------------------------------------------------------------
// +++ Observation fusion +++
// -------------------------------------------------------------------------------
if(CIWTApp::debug_level>0) printf("->Observation fusion ...\r\n");
std::vector<GOT::tracking::Observation> observations_all;
const auto &proposal_set_to_use = object_proposals_all;
if (CIWTApp::tracking_mode_str=="detection") {
/// Baseline: use only what can be inferred from detection bounding-boxes (faster, less accurate in 3D).
auto det_obs_fnc = GOT::tracking::observation_processing::DetectionsOnly;
observations_all = det_obs_fnc(detections_current_frame, proposal_set_to_use, left_image);
}
else if (CIWTApp::tracking_mode_str=="detection_shape") {
/// Do proposal<->detection association using CRF, for precise detection localization.
observations_all = GOT::tracking::observation_processing::DetectionToSegmentationFusion(
detections_current_frame, proposal_set_to_use,
left_camera, left_image, variables_map
);
}
else {
std::cout << "Unknown tracking_mode: " + CIWTApp::tracking_mode_str + "." << std::endl;
return 0;
}
// -------------------------------------------------------------------------------
// +++ Re-compute pos. cov. matrices +++
// -------------------------------------------------------------------------------
// Add some (learned) 'bias' to covariance matrices, depending on wheter it was localized with an 3D segment or by footpoint-projection.
if(CIWTApp::debug_level>0) printf("->Override cov-mats ...\r\n");
for (auto &obs:observations_all) {
Eigen::Matrix3d cov_out;
left_camera.ComputeMeasurementCovariance3d(obs.footpoint().head<3>(), 0.5, left_camera.P().block(0,0,3,4), right_camera.P().block(0,0,3,4), cov_out);
double add_unc_x = 0.0;
double add_unc_z = 0.0;
if (obs.proposal_3d_avalible()) {
add_unc_x = variables_map["pos_unc_seg_x"].as<double>();
add_unc_z = variables_map["pos_unc_seg_z"].as<double>();
} else {
add_unc_x = variables_map["pos_unc_det_x"].as<double>();
add_unc_z = variables_map["pos_unc_det_z"].as<double>();
}
cov_out(0, 0) += add_unc_x;
cov_out(2, 2) += add_unc_z;
obs.set_covariance3d(cov_out);
}
// -------------------------------------------------------------------------------
// +++ Compute velocity for segments +++
// -------------------------------------------------------------------------------
if (dataset_assistant.velocity_map_.data!=nullptr) {
// If velocity map is provided, we can compute velocities for fused observations
if(CIWTApp::debug_level>0) printf("->Compute segment velocities ...\r\n");
for (auto &obs:observations_all) {
if (obs.proposal_3d_avalible()) {
Eigen::Vector3d obs_velocity = SUN::utils::observations::ComputeVelocity(velocity_map, obs.pointcloud_indices(),
variables_map["dt"].as<double>());
obs.set_velocity(obs_velocity);
}
}
}
const int max_num_observations_for_tracker = variables_map["max_num_input_observations"].as<int>();
const int num_best_observations_to_consider = std::min(max_num_observations_for_tracker, static_cast<int>(observations_all.size()));
auto observations_to_pass_to_tracker = observations_all;
// -------------------------------------------------------------------------------
// +++ Tracking +++
// -------------------------------------------------------------------------------
/// Feed the resource manager with current-frame data
resource_manager->AddNewMeasurements(current_frame, left_point_cloud_preprocessed, left_camera);
resource_manager->AddNewObservations(observations_to_pass_to_tracker);
/// Invoke the tracker
if (CIWTApp::run_tracker) {
if(CIWTApp::debug_level>0) printf("->Running the tracker (observations: %d) ...\r\n", (int)observations_to_pass_to_tracker.size());
multi_object_tracker->ProcessFrame(resource_manager, current_frame);
}
// Timing analysis
std::chrono::steady_clock::time_point time_end = std::chrono::steady_clock::now();
total_processing_time += std::chrono::duration_cast<std::chrono::milliseconds>(time_end - time_begin).count();
auto hypos_to_process_further = multi_object_tracker->selected_hypotheses(); // Selected hypothesis set.
auto hypos_terminated = multi_object_tracker->terminated_hypotheses(); // Terminated, but still active, set.
/// Export tracking data to labels struct (need for results export)
GOT::tracking::utils::HypothesisSetToLabels(current_frame, hypos_to_process_further, tracker_result_labels,
std::bind(GOT::tracking::utils::HypoToLabelDefault, std::placeholders::_1, std::placeholders::_2,
left_camera.width(), left_camera.height()));
// -------------------------------------------------------------------------------
// +++ Visualizations (image) +++
// -------------------------------------------------------------------------------
/// Draw observations and tracked objects
cv::Mat left_image_with_observations = left_image.clone();
cv::Mat left_image_with_hypos_2d = left_image.clone();
cv::Mat left_image_with_detections = left_image.clone();
if (CIWTApp::debug_level>=2 || CIWTApp::show_visualization_2d) {
CIWTApp::tracking_visualizer.DrawObservations(observations_all, left_image_with_observations, left_camera, GOT::tracking::draw_observations::DrawObservationAndOrientation);
CIWTApp::tracking_visualizer.DrawHypotheses(hypos_to_process_further, left_camera, left_image_with_hypos_2d, GOT::tracking::draw_hypos::DrawHypothesis2d);
}
if (CIWTApp::show_visualization_2d) {
cv::imshow("tracking_2d_window", left_image_with_hypos_2d);
cv::imshow("observations_id_window", left_image_with_observations);
}
/// Save to the disk
if (CIWTApp::debug_level>=2) {
char frame_str_buff[50];
snprintf(frame_str_buff, 50, (CIWTApp::sequence_name + "_%06d").c_str(), current_frame);
char output_path_buff[500];
if (CIWTApp::run_tracker) {
// Tracking 2D visualization
snprintf(output_path_buff, 500, "%s/hypos_2d_%s.png", output_dir_visual_results.c_str(), frame_str_buff);
cv::imwrite(output_path_buff, left_image_with_hypos_2d);
}
}
// -------------------------------------------------------------------------------
// +++ Update the state for the 3D visualizer thread +++
// -------------------------------------------------------------------------------
if (CIWTApp::show_visualization_3d) {
boost::mutex::scoped_lock updateLock(CIWTApp::visualization_3d_update_mutex);
CIWTApp::visualization_3d_update_flag = true;
pcl::copyPointCloud(*dataset_assistant.left_point_cloud_, *CIWTApp::visualization_3d_point_cloud);
CIWTApp::visualization_3d_point_cloud->sensor_origin_ = Eigen::Vector4f(0.0,0.0,0.0,1.0);
CIWTApp::visualization_observations = observations_all;
CIWTApp::visualization_3d_proposals = proposal_set_to_use;
char frame_str_buff[50];
snprintf(frame_str_buff, 50, (CIWTApp::sequence_name).c_str());
char buff[500];
snprintf(buff, 500, "%s/3d_viewer_%s_%06d.png", output_dir_visual_results.c_str(), frame_str_buff, current_frame);
CIWTApp::viewer_3D_output_path = std::string(buff);
CIWTApp::visualization_3d_tracking_hypotheses = hypos_to_process_further;
CIWTApp::visualization_3d_tracking_terminated_hypotheses = hypos_terminated;
CIWTApp::visualization_3d_left_camera = left_camera;
updateLock.unlock();
}
}
// -------------------------------------------------------------------------------
// +++ END-OF MAIN TRACKING LOOP +++
// -------------------------------------------------------------------------------
printf ("================ PROCESSING TIME ================\n");
printf ("Total processing time %f (milisec) for %d frames.\n", total_processing_time, CIWTApp::end_frame-CIWTApp::start_frame);
printf ("That is %f (milisec) per-frame.\n", total_processing_time / static_cast<double>(CIWTApp::end_frame-CIWTApp::start_frame));
printf ("Note: processing time excludes tracker-input-data processing (disparity estimation, proposal generation etc.)\n");
printf ("=================================================\n");
// KITTI I/O label export obj.
SUN::utils::KITTI::LabelsIO kitti_tracking_output_io;
/// Write-out tracking quantitative results
std::string tracking_output_dir = CIWTApp::output_dir + "/data";
SUN::utils::IO::MakeDir(tracking_output_dir.c_str());
char hypothesis_file_in_kitti_format[500];
snprintf(hypothesis_file_in_kitti_format, 500, "%s/%s.txt", tracking_output_dir.c_str(), CIWTApp::sequence_name.c_str());
kitti_tracking_output_io.WriteLabels(tracker_result_labels, hypothesis_file_in_kitti_format);
std::cout << "Finished, yay!" << std::endl;
return 0;
}