-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathindex.html
More file actions
1007 lines (741 loc) · 47.7 KB
/
index.html
File metadata and controls
1007 lines (741 loc) · 47.7 KB
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
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!doctype html>
<html>
<head>
<link rel="stylesheet" href="common/css/reveal.css">
<link rel="stylesheet" href="common/css/white.css">
<link rel="stylesheet" href="common/css/overrides.css">
<script src="common/js/print.js"></script>
<title>A Pipeline for Distributed High-resolution 3D Reconstruction of Urban Scenes from Point Clouds</title>
<meta charset="UTF-8">
</head>
<body>
<div class="reveal">
<div class="slides">
<!-- ***
* Title Slide Branded
* -->
<section data-transition="fade" data-background-image="common/assets/img/titlePage.png">
<div style="position: absolute; top: 0px; left: 0px;">
<div style="color: black; margin-top: 0px; text-align: left; margin-left: 340px; margin-top: 40px; line-height: 100%">
<div style="font-size: 100%;"><strong>A Pipeline for Distributed High-resolution 3D Reconstruction of Urban Scenes from Point Clouds</strong></div>
<div style="font-size: 70%; margin-top: 42px;">
May 18 2021 9:45 AM - IEEE STRATUS 2021
</div>
</div>
<div style="color: white; margin-left: 340px; margin-top: 180px; font-size: 55%;">
<div style="text-align: left; float: left;">Ty Feng<span style="font-size:80%"> - Department of Computer Science - Institute for Data Science and Computing - <a href="mailto:artyfeng@miami.edu" style="color: #fff; text-decoration: underline;">artyfeng@miami.edu</a></span></div>
<div style="text-align: left; float: left;">
Amin Sarafraz<span style="font-size:80%"> - Department of Civil, Architectural and Environmental Engineering - Institute for Data Science and Computing</span><br>
Christopher Mader<span style="font-size:80%"> - Institute for Data Science and Computing</span>
</div>
<div style="font-size: 70%; margin-top: 70px; text-align: left; float: left;">
<a href="https://ytyfeng.github.io/Distributed_Recon/">Slides available on ytyfeng.github.io/Distributed_Recon/</a></li>
</div>
</div>
</div>
</section>
<!-- ***
* Outline
* -->
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span></span></div>
<div class="template-content-containter">
<div class="template-title"><span>Outline<span></div>
<div class="template-content">
<ul style="line-height: 180%;">
<li>Introduction</li>
<li>Prior Work</li>
<li>Our Approach</li>
<li>Implementation</li>
<li>Results</li>
</ul>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li></li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<!-- ***
* Proposal
* -->
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>the problem</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Introduction<span></div>
<div class="template-content">
With the progress in aerial 3D acquisition technologies such as camera drones and aerial LiDAR,
it is now possible to have 3D scans of large urban scenes.
The result of such scans is registered and aligned to the same coordinate system
and then further processed and reconstructed to 3D mesh models. However, processing
and reconstructing gigabytes of large-scale point cloud data with complex real-world geometry is a challenging task.
It requires significant computing powers and memory usage that typical personal computers
cannot afford. Aside from the memory challenges, processing and reconstructing large point clouds
requires significant amount of computing time.
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li></li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade" data-background-video="common/assets/video/drone_flight.mp4" data-background-video-loop data-background-video-muted></section>
<!-- ***
* Building a mesh from point cloud
* -->
<section data-transition="fade" data-background-image="common/assets/img/pointcloud.jpg"></section>
<section data-transition="fade" data-background-image="common/assets/img/wireframe.jpg"></section>
<section data-transition="fade" data-background-image="common/assets/img/blend.jpg"></section>
<section data-transition="fade" data-background-image="common/assets/img/mesh.jpg"></section>
<!-- ***
* prior work
* -->
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>prior work</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Existing parallel methods<span></div>
<div class="template-content" style="text-align: left;">
<ul>
<li>
Multiple threads on a single node:
<ul>
<li>Can save time but limited by memory capacity of the node</li>
</ul>
</li>
<li>
Out-of-core methods:
<ul>
<li>Can store more data into disk and swap into memory when needed, but time is bottlenecked by computer's I/O speed</li>
</ul>
</li>
</ul>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li>Julie Digne. An Analysis and Implementation of a Parallel Ball Pivoting Algorithm.Image Processing On Line, 4:149–168, 2014</li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>prior work</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Surface Reconstruction<span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
<ul style="margin-top: 3%;">
<li>Explicit, geometric reconstruction</li>
<ul>
<li>Ball-pivoting algorithm</li>
<ul><li>Radius selection problem, holes, require uniformly sampled point clouds</li> </ul>
<li>Delaunay-based triangulation</li>
<ul><li>Provably correct, but require densely-sampled, noise-free point clouds</li></ul>
</ul>
</ul>
</ul>
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li>Fausto Bernardini, Joshua Mittleman, Holly Rushmeier, Claudio Silva, and Gabriel Taubin. The ball-pivoting algorithm for surface reconstruction. IEEE Transactions on Visualization and Computer Graphics, 5(4):349–359, Oct 1999.</li>
<li>Frédéric Cazals and Joachim Giese n.Delaunay Triangulation Based Surface Reconstruction, pages 231–276. Springer Berlin Heidelberg, Berlin, Heidelberg, 2006.</li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>prior work</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Surface Reconstruction<span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
<ul style="margin-top: 3%;">
<li>Implicit, function-fitting reconstruction</li>
<ul>
<li>Poisson Reconstruction</li>
<ul>
<li>
Fit an implicit function to indicator gradient from oriented points (point + normal)
</li>
<li> Indicator gradient: 0 almost everywhere, 1 at surface; since indicator gradient is equal to inward surface normal (normalized)</li>
<li>Adaptive octree to represent the implicit function and to solve the Poisson problem. Increasing octree max depth increases mesh resolution (triangles and file size), time, and memory by a factor of 4</li>
<li>Extract 0-th level isosurface from implicit function using Marching Cubes</li>
</ul>
</ul>
</ul>
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li>Michael Kazhdan, Matthew Bolitho, and Hugues Hoppe. Poisson surface reconstruction. In Proceedings of the Fourth Eurographics Symposium on Geometry Processing, SGP 06, page 6170, Goslar, DEU, 2006. Eurographics Association.</li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>prior work</span></div>
<div class="template-content-containter">
<div class="template-title"><span>BPA vs Poisson<span></div>
<div class="template-content" style="max-width: 90% !important;">
<img src="common/assets/img/bpa_poisson.png">
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li><a href="http://tyfeng.com/PoissonVisualizer">tyfeng.com/PoissonVisualizer</a></li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<iframe style="width:100%; height:700px" src="https://ocf.berkeley.edu/~tyfeng/PoissonVisualizer"> </iframe>
<div class="template-citation">
<ul style="list-style: none;">
<li><a href="http://tyfeng.com/PoissonVisualizer">tyfeng.com/PoissonVisualizer</a></li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<!--
our method
-->
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>our method</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Our method<span></div>
<div class="template-content" style="max-width: 90% !important;">
Fully independent, distributed, multi-node method
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li> </li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>our method</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Our reconstruction pipeline<span></div>
<div class="template-content" style="max-width: 90% !important;">
<img src="common/assets/img/pipeline.png">
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li> </li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>our method</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Our reconstruction pipeline<span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
<ul style="margin-top: 3%;">
<li>Chunk Point Clouds</li>
<ul>
<li>
Chunk point clouds by x and y axes since the range of z values
are relatively small in urban scenes
</li>
<li>Compute the axis-aligned bounding box (max and min bounds)</li>
<li>Compute an x by y grid for point cloud's bounding box</li>
<li>Crop by x * y bounding boxes</li>
</ul>
<li>Adding overlap</li>
<ul>
<li>Degraded reconstruction along edges (open shapes)</li>
<li>Add a 5% overlap before meshing</li>
</ul>
</ul>
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li></li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>our method</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Our reconstruction pipeline<span></div>
<div class="template-content" style="max-width: 90% !important;">
<figure>
<img src="common/assets/img/chunks.png" style="width:30%">
<figcaption>
Adding overlapping boundary to each chunk. Blue boxes represent original bounding boxes.
Magenta boxes represent bounding boxes with overlapping boundary.
</figcaption>
</figure>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li> </li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>our method</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Our reconstruction pipeline<span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
<ul style="margin-top: 3%;">
<li>Fully-independent, distributed, multi-node processing</li>
<ul>
<li>After chunking a point cloud into manageable chunks, we distribute the subsequent processing and meshing of each chunk as an independent process on different nodes on an HPC cluster</li>
<li>Fully-independent processes because chunking fully separates data; the failure of one process does not have an impact on other processes</li>
<li>Memory usage is bound by the capacity of a node, but chunking and distributed processing on multiple nodes reduce memory needed for each node</li>
<li>Each chunk of point cloud is processed and meshed with a single process from start to finish, so it is easier to debug by simply reading the output and error log associated with a particular point cloud chunk</li>
</ul>
</ul>
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li></li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>our method</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Our reconstruction pipeline<span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
<ul style="margin-top: 3%;">
<li>Normal estimation</li>
<ul>
<li>Nearest neighbor search within a search radius and/or k-neighbors</li>
<li>Principal component analysis (PCA) finds an orthogonal basis that best represents a given neighborhood of points. It finds the best-fitting plane by minimizing the average squared distance from a neighborhood point to the plane </li>
<li>The normal is a unit vector orthogonal to the plane</li>
</ul>
</ul>
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li>Hugues Hoppe, Tony DeRose, Tom Duchamp, John McDonald, and Werner Stuetzle. Surface reconstruction from unorganized points. SIGGRAPH Comput. Graph.,26(2):7178, July 1992.</li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>our method</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Our reconstruction pipeline<span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
<ul style="margin-top: 3%;">
<li>Poisson reconstruction</li>
<ul>
<li>Point cloud with normals can be meshed using Poisson Reconstruction with a user-defined octree depth</li>
</ul>
<li>Merging</li>
<ul>
<li>Using original point cloud bounding boxes, crop the overlapped meshes back</li>
<li>Add 1% overlap if necessary</li>
</ul>
</ul>
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li></li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>our method</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Our reconstruction pipeline<span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
<ul style="margin-top: 3%;">
<li>Reconstruction at higher resolution</li>
<ul>
<li>
We can offset the increased memory usage due to increase in resolution/octree depth by reducing the chunk size.
</li>
<li>
Increasing octree depth increases time, memory usage, and number of triangles by roughly a factor of 4.
</li>
<table>
<thead>
<tr>
<th>Depth</th>
<th>Time</th>
<th>Maximum Memory</th>
<th>Triangles</th>
</tr>
</thead>
<tbody>
<tr>
<td>8</td>
<td>141.46 sec</td>
<td>243 MB</td>
<td>202,163</td>
</tr>
<tr>
<td>9</td>
<td>234.90 sec</td>
<td>629 MB</td>
<td>957,104</td>
</tr>
<tr>
<td>10</td>
<td>677.01 sec</td>
<td>1980 MB</td>
<td>4,524,372</td>
</tr>
<tr>
<td>11</td>
<td>2663.84 sec</td>
<td>6646 MB</td>
<td>20,857,780</td>
</tr>
<tr>
<td>12</td>
<td>10985.97 sec</td>
<td>23781 MB</td>
<td>97,923,718</td>
</tr>
</tbody>
</table>
</ul>
</ul>
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>our method</span></div>
<div class="template-content-containter">
<div class="template-title"><span> <span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
<img src="common/assets/img/cars.png" />
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li>
Comparison of Poisson meshes generated using different octree depth values (Data: cars in a parking lot)
</li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<!-- ***
* Implementation
* -->
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>implementation</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Implementation<span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
<ul>
<li>Implemented on IBM Power9 HPC cluster (ppcle64)</li>
<li>Shell commands for each process > shell script</li>
<li>Submit shell script as a job to LSF job scheduler on cluster</li>
<li>
Each process does the following for a point cloud chunk:
<ul style="font-size: 90%;">
<li>Adding a 5% overlap</li>
<li>kNN and PCA for normal estimation</li>
<li>Poisson Reconstruction</li>
<li>Cropping back to original bounding box</li>
</ul>
<li>We used <a href="http://open3d.org">Open3D </a> </li>
</ul>
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li>Qian-Yi Zhou, Jaesik Park, and Vladlen Koltun. Open3D: A modern library for 3D data processing. arXiv:1801.09847, 2018.</li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<!--***
* Results
* -->
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>results</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Results<span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
<table>
<thead>
<tr>
<th>Method</th>
<th>Total time (seconds)</th>
<th>Memory usage per node (GB)</th>
</tr>
</thead>
<tbody>
<tr>
<td>Single node without chunking</td>
<td>2233</td>
<td>250.341</td>
</tr>
<tr>
<td>
Our method with
36 chunks (15 nodes)
</td>
<td>955</td>
<td>16.972</td>
</tr>
<tr>
<td>
Our method with
36 chunks (30 nodes)
</td>
<td>904</td>
<td>8.185</td>
</tr>
</tbody>
</table>
Max time and memory usage for processing and reconstructing 1.18 billion points
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>results</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Results<span></div>
<div class="template-content" style="max-width: 100% !important;">
<img src="common/assets/img/36_chunks.png">
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li> Result of our reconstruction with example data divided into 36 chunks</li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>conclusion</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Conclusion<span></div>
<div class="template-content" style="max-width: 85% !important;">
We have shown that large-scale 3D reconstruction can be handled using a distributed method on
computers with memory capacity similar to personal computers’ by first chunking the point cloud
and then processing the chunks on multiple nodes. We have demonstrated that large memory is not required
to process and mesh large-scale point clouds. By addressing the memory issue using our method, we can
also greatly improve the reconstruction resolution using higher octree depth levels.
The resulting mesh is high in resolution and show no negative effects due to open shapes caused by
our chunking process.
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>credits</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Acknowledgments<span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
University of Miami Institute for Data Science and Computing, Advanced Computing
<br>
for the use of HPC clusters
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li></li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>credits</span></div>
<div class="template-content-containter">
<div class="template-title"><span>References<span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
<ul>
<li>Julie Digne. An Analysis and Implementation of a Parallel Ball Pivoting Algorithm. Image Processing On Line, 4:149–168, 2014</li>
<li>Fausto Bernardini, Joshua Mittleman, Holly Rushmeier, Claudio Silva, and Gabriel Taubin. The ball-pivoting algorithm for surface reconstruction. IEEE Transactions on Visualization and Computer Graphics, 5(4):349–359, Oct 1999.</li>
<li>Frédéric Cazals and Joachim Giese n.Delaunay Triangulation Based Surface Reconstruction, pages 231–276. Springer Berlin Heidelberg, Berlin, Heidelberg, 2006.</li>
<li>Michael Kazhdan, Matthew Bolitho, and Hugues Hoppe. Poisson surface reconstruction. In Proceedings of the Fourth Eurographics Symposium on Geometry Processing, SGP 06, page 6170, Goslar, DEU, 2006. Eurographics Association.</li>
</ul>
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li></li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>credits</span></div>
<div class="template-content-containter">
<div class="template-title"><span>References<span></div>
<div class="template-content" style="max-width: 90% !important;">
<div style="display: inline-block; width: 100%">
<ul>
<li>William E. Lorensen and Harvey E. Cline. Marching cubes: A high resolution 3d surface construction algorithm. In Proceedings of the 14th Annual Conference on Computer Graphics and Interactive Techniques, SIGGRAPH 87, page 163169, New York, NY, USA, 1987. Association for Computing Machinery.</li>
<li>Hugues Hoppe, Tony DeRose, Tom Duchamp, John McDonald, and Werner Stuetzle. Surface reconstruction from unorganized points. SIGGRAPH Comput. Graph.,26(2):7178, July 1992.</li>
<li>Qian-Yi Zhou, Jaesik Park, and Vladlen Koltun. Open3D: A modern library for 3D data processing. arXiv:1801.09847, 2018.</li>
<li>Michael Kazhdan. Adaptive multigrid solvers (version 13.00), June 2020.</li>
</ul>
</div>
</div>
</div>
<div class="template-citation">
<ul style="list-style: none;">
<li></li>
</ul>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
<section data-transition="fade" data-background-video="common/assets/video/ML_onetile.mp4" data-background-video-loop data-background-video-muted>
<div>
<ul style="list-style: none; padding-top:600px;">
<li>Video credit: Timothy Norris (University of Miami, Tiling the MagicVerse) </li>
</ul>
</div>
</section>
<section data-transition="fade">
<div class="template-container">
<div class="template-header"><span>thanks</span></div>
<div class="template-content-containter">
<div class="template-title"><span>Thank you<span></div>
<div class="template-content" style="max-width: 90% !important;">
Slides available on <a href="https://ytyfeng.github.io/Distributed_Recon/">https://ytyfeng.github.io/Distributed_Recon/</a></li>
</div>
</div>
<div class="template-footer"><table width=100%><tr><td background="common/assets/img/UM.jpg" height=80 width=98></td><td background="common/assets/img/bar.jpg"></td><td width=60></td></tr></table></div>
</div>
</section>
</div>
</div>
<script src="common/js/reveal.js"></script>
<script src="common/js/notes.js"></script>
<script src="common/js/head.min.js"></script>
<script>Reveal.initialize({
width: 1366, height: 768, margin: 0.05,
controls: false,
progress: true,
history: true,
center: true,
mouseWheel: false,
chart: {
defaults: {
global: {
defaultFontSize: 20,
title: { fontColor: "#333" },
legend: {
position: "top",
labels: { fontColor: "#333" },
},
tooltips: {
labels: { fontColor: "#333" },
},
},
scale: {
scaleLabel: { fontColor: "#333" },
gridLines: { color: "#333", zeroLineColor: "#333" },
ticks: { fontColor: "#333" },
}
},
},