-
Notifications
You must be signed in to change notification settings - Fork 3
/
index.html
614 lines (532 loc) · 28.8 KB
/
index.html
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
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta name="description" content="">
<meta name="author" content="">
<title>Unicorn - Landing Page</title>
<!-- Bootstrap core CSS -->
<link href="css/bootstrap.min.css" rel="stylesheet">
<!-- Animation CSS -->
<link href="css/animate.min.css" rel="stylesheet">
<link href="css/font-awesome.min.css" rel="stylesheet">
<!-- HTML5 shim and Respond.js for IE8 support of HTML5 elements and media queries -->
<!--[if lt IE 9]>
<script src="https://oss.maxcdn.com/html5shiv/3.7.2/html5shiv.min.js"></script>
<script src="https://oss.maxcdn.com/respond/1.4.2/respond.min.js"></script>
<![endif]-->
<!-- Custom styles for this template -->
<link href="css/style.css" rel="stylesheet">
</head>
<body id="page-top">
<div class="navbar-wrapper">
<nav class="navbar navbar-default navbar-fixed-top" role="navigation">
<div class="container">
<div class="navbar-header page-scroll">
<button type="button" class="navbar-toggle collapsed" data-toggle="collapse" data-target="#navbar" aria-expanded="false" aria-controls="navbar">
<span class="sr-only">Toggle navigation</span>
<span class="icon-bar"></span>
<span class="icon-bar"></span>
<span class="icon-bar"></span>
</button>
<a class="navbar-brand" href="#">Unicorn</a>
</div>
<div id="navbar" class="navbar-collapse collapse">
<ul class="nav navbar-nav navbar-right">
<li><a class="page-scroll" href="#page-top">Home</a></li>
<li><a class="page-scroll" href="#intro">Introduction</a></li>
<li><a class="page-scroll" href="#method">Method</a></li>
<li><a class="page-scroll" href="#exp">Experiments</a></li>
<!-- <li><a class="page-scroll" href="#introduction">Details</a></li>
<li><a class="page-scroll" href="#codes">Codes</a></li> -->
<li><a class="page-scroll" href="#team">Team</a></li>
</ul>
</div>
</div>
</nav>
</div>
<div id="inSlider" class="carousel slide carousel-fade" data-ride="carousel">
<!-- <ol class="carousel-indicators">
<li data-target="#inSlider" data-slide-to="0" class="active"></li>
</ol> -->
<div class="carousel-inner" role="listbox">
<div class="item active">
<div class="container">
<div class="carousel-caption">
<div class="col-lg-12 text-center">
<img style="width:7%;" src="./img/unicorn-logo1.png">  
<img style="width:18%;" src="./img/unicorn-logo2.png">
<h1>A Versatile Point Cloud Compressor<br/>
Using Universal Multiscale Conditional Coding:<br/>
Geoemetry & Attribute
</h1>
<!-- <h1 style="font-size:30px"> TPAMI 2024</h1> -->
<!-- <div class="is-size-5 publication-authors">
<span style="font-size:16px">
Jianqiang Wang<sup># 1</sup>
</span>
<span style="font-size:16px">
Ruixiang Xue<sup># 1</sup></span>
<span style="font-size:16px">
Jiaxin Li<sup>1</sup>
</span>
<span style="font-size:16px">
Dandan Ding<sup>2</sup>
</span>
<span style="font-size:16px">
Yi Lin<sup>* 3</sup>
</span>
<span style="font-size:16px">
Zhan Ma<sup>* 1</sup>
</span>
</div>
<div style="font-size:16px">
<sup># </sup> Denotes Equal Contribution
<sup>* </sup> Denotes Corresponding Authors
</div> -->
<br>
<div class="is-size-5 publication-authors">
<span style="font-size:24px" ><sup>1</sup> Nanjing University <a href="https://vision.nju.edu.cn/"> Vision Lab </a></span>
<span style="font-size:24px"><sup>2</sup>Hangzhou Normal University</span>
<span style="font-size:24px"><sup>3</sup>Fudan University</span>
</div>
<!-- <div style="font-size:24px">
<sup>1</sup> <a href="https://vision.nju.edu.cn/main.htm"><img style="width:5%;" src="./img/nju.jpg"> </a>
<sup>2</sup> <a href="https://vision.nju.edu.cn/main.htm"><img style="width:5%;" src="./img/hznu.jpg"> </a>
<sup>3</sup> <a href="https://vision.nju.edu.cn/main.htm"><img style="width:5%;" src="./img/fudan.jpeg"> </a>
</div> -->
<br>
<a href="https://ieeexplore.ieee.org/document/10682571">
<span style="font-size:20px" >Paper I </span>
</a>
<a href="https://ieeexplore.ieee.org/document/10682566">
<span style="font-size:20px" >Paper II </span>
</a>
<a href="https://github.com/NJUVISION/Unicorn">
<span style="font-size:20px" >Code </span>
</a>
<!-- <a href="https://github.com/NJUVISION/Unicorn">
<span style="font-size:20px" >Datasets </span>
</a> -->
<a href="./results.zip">
<span style="font-size:20px" >Results </span>
</a>
<a href="./proposal.zip">
<span style="font-size:20px" >Proposals </span>
</a>
</div>
</div>
</div>
</div>
</div>
</div>
<section id="1", class="results">
<div class="container">
<div class="row">
<div class="navy-line"></div>
<div class="col-lg-12 text-center">
<h2> <b>News</b> </h2>
</div>
<div class="col-lg-12 text-left" style="font-size:20px">
<strong> 2024.12.06 Open source Unicorn Pre (<a href="https://github.com/NJUVISION/SparsePCGC">SparsePCGC</a>)!
<br><br>
<strong> 2024.10.28 Unicorn version 2 has responded to the Call for Proposals for AI-based Point
Cloud Coding (m70061 & m70062 in <a href="https://dms.mpeg.expert/">MPEG</a>).</strong>
<br><br>
<strong> 2024.10.05 Initial release of part of the code and results. (The entire source code will be
released to the public after the approval from the funding agency.)</strong>
<br><br>
<strong> 2024.09.12 Unicorn version 1 was accepted by TPAMI.
(<a href="https://ieeexplore.ieee.org/document/10682571"> Part I</a> and
<a href="https://ieeexplore.ieee.org/document/10682566"> Part II </a>)</strong>
</div>
</div>
</div>
</section>
<section id="intro", class="results">
<div class="container">
<div class="row">
<div class="navy-line"></div>
<div class="col-lg-12 text-center">
<h2> <b>Abstract</b> </h2>
</div>
<div class="col-lg-12 text-left" style="font-size:20px">
A universal multiscale conditional coding framework, <b>Unicorn</b>,
is proposed to compress the geometry and attribute of any given point cloud.
Geometry compression is addressed in <a href="https://ieeexplore.ieee.org/document/10682571"> Part I</a> of this paper,
while attribute compression is discussed in <a href="https://ieeexplore.ieee.org/document/10682566"> Part II</a>.
<br><br>
For <b>geoemtry </b> compression, we construct the multiscale sparse tensors
of each voxelized point cloud frame
and properly leverage lower-scale priors in the current
and (previously processed) temporal reference frames
to improve the conditional probability approximation
or content-aware predictive reconstruction of geometry occupancy in compression.
<br><br>
For <b>attribute</b> compression, Since attribute components exhibit
very different intrinsic characteristics
from the geometry element, e.g., 8-bit RGB color versus 1-bit occupancy,
we process the attribute residual between lower-scale reconstruction
and current-scale data.
Similarly, we leverage spatially lower-scale priors in the current frame and
(previously processed) temporal reference frame to improve the probability
estimation of attribute intensity through conditional residual prediction
in lossless mode or enhance the attribute reconstruction through
progressive residual refinement in lossy mode for better performance.
<br><br>
The porposed Unicorn is a versatile,
learning-based solution capable of compressing
static and dynamic point clouds with diverse source characteristics
in both lossy and lossless modes. Following the same evaluation criteria,
Unicorn significantly outperforms standard-compliant approaches
like MPEG G-PCC, V-PCC, and other learning-based solutions, yielding
state-of-the-art compression efficiency while presenting
affordable complexity for practical implementations.
</div>
<!-- <div class="col-lg-12 text-center" style="font-size:20px">
<br>
<h2> Video </h2>
<video width="960" height="580" controls="controls">
<source src="img/v1.mp4" type="video/mp4" />
Your browser does not support the video tag.
</video>
</div> -->
<div class="col-lg-12 text-center" style="font-size:20px">
<h2> <b>Contributions</b> </h2>
</div>
<div class="col-lg-12 text-left" style="font-size:20px">
<b>Comprehensive coding metric:</b> 
Unicorn is the first, versatile, learning-based PCC solution.
<br> 1) It can compress the geometry and attribute information, either separately or jointly, of an input point cloud.
<br> 2) It flexibly supports the static and dynamic coding of point clouds in either lossless or lossy mode.
<br> 3) It demonstrates the leading performance for diverse types, including solid, dense, and sparse object point clouds,
as well as scant LiDAR.
</div>
<div class="col-lg-12 text-left" style="font-size:20px">
<b>Better compression performance: </b> 
Unicorn provides significant performance gains to existing approaches.
</div>
<div class="col-lg-12 text-left" style="font-size:20px">
<b>Low computation complexity: </b> 
Unicorn is a low-complexity approach with comparable runtime measures to the G-PCC codec
and variable-rate coding capability using a single neural model.
</div>
</div>
</div>
</section>
<section id="method", class="results">
<div class="container">
<div class="row">
<div class="col-lg-12 text-center">
<div class="navy-line"></div>
<h2> <b>Method</b> </h2>
<img src="img/unicorn-data.png" style="width:80%;">
<h4> <b>Data processing in Unicorn. </b>
A specific frame P<sub>tk</sub> includes the geometry part O<sub>tk</sub>
and attribute intensity I<sub>tk</sub>;
Voxelized O<sub>tk</sub> is represented using sparse tensor
that only contains occupied voxels.
</h4>
<br><br>
<div class="text-center" style="font-size:18px">
<h2>Geometry
        
        
        
       Attribute</h2>
</div>
<!-- <div class="text-center">
<h2>Attribute</h2>
</div> -->
<br>
<img src="img/MST_geo.png" style="width:40%;">           
<img src="img/MST_attr.png" style="width:40%;">
<h4> <b>Unicorn's Multiscale Sparse Representation. </b>
(left) geometry: 1 - Occupied voxel, 0 - Unoccupied voxel;
(right) color attribute exemplified using luma or Y intensity.
OPU is the Occupancy Processing Unit,
and APU is the Attribute Processing Unit.
</h4>
<br><br>
<img src="img/OPU.png" style="width:40%;">           
<img src="img/APU.png" style="width:40%;">
<h4> <b>Cross-scale Processing Units. </b>
(left) OPU; (right) APU.
Spatially or spatiotemporally lower-scale priors
are used to support probability approximation in lossless mode
or predictive/progressive reconstruction in lossy mode
for respective static or dynamic coding.
</h4>
<br><br>
<img src="img/lossless_geometry.png" style="width:40%;">
              
<img src="img/lossless_attribute.png" style="width:40%;">
<h4> <b>Losslesss Coder in Unicorn. </b>
(left) geometry; (right) attribute.
</h4>
<br>
<img src="img/lossy_geometry.png" style="width:40%;">
              
<img src="img/lossy_attribute.png" style="width:40%;">
<h4> <b>Lossy Coder in Unicorn. </b>
(left) geometry; (right) attribute.
</h4>
<br>
<img src="img/dynamic_geometry.png" style="width:40%;">
              
<img src="img/dynamic_attribute.png" style="width:40%;">
<h4> <b>Dynamic Coder in Unicorn. </b>
(left) geometry; (right) attribute.
</h4>
<br>
<img src="img/IntegratedFramework.png" style="width:55%;">
<h4> <b>Unified compression of geometry and attribute in Unicorn. </b>
</h4>
</div>
</div>
</div>
</section>
<section id="exp", class="results"></section>
<div class="container">
<div class="row">
<div class="col-lg-12 text-center">
<div class="navy-line"></div>
<h2> <b>Experiments</b> </h2>
<br>
<img src="img/dataset.png" style="width:70%;">    
<!-- <br> -->
<!-- <video width="250" height="250" class="center" controls="controls">
<source src="img/longdress.mp4" type="video/mp4"/>
</video> -->
<!-- <video width="250" height="250" controls="controls">
<source src="img/lidar.mp4" type="video/mp4"/>
</video> -->
<h4> <b>Point Cloud Examples. </b>
We conducted extensive experiments
on various point cloud datasets
to thoroughly understand the efficiency and generalization
of Unicorn.
These datasets include static and dynamic samples with diverse contents,
densities, resolutions, and other characteristics.
</h4>
<br>
<div class="text-center" style="font-size:16px"></div>
<h2>Part I: Geometry </h2>
<br>
<img src="img/geometry/SolidObjectvox108iVFBOwliiD1PSNR.png" style="width:30%;">  
<img src="img/geometry/DenseObjectvox12FacadeHouseD1PSNR.png" style="width:30%;">  
<img src="img/geometry/SparseObjectvox12AcroShivaStaueD1PSNR.png" style="width:30%;">  
<h4> <b>R-D comparison for static geometry coding. </b>
</h4>
<br><br>
<img src="img/geometry/SolidObjectvox10OwliimseFPSNRp2point.png" style="width:30%;">  
<img src="img/geometry/ScantLiDARSceneq1mmFordD1PSNR.png" style="width:30%;">  
<img src="img/geometry/ScantLiDARSceneq1mmKITTID1PSNR.png" style="width:30%;">  
<h4> <b>R-D comparison for dynamic geometry coding. </b>
</h4>
<br><br>
<img src="img/ablation_error_map.png" style="width:80%;">  
<h4> <b>Error map visualization of reconstructed point clouds. </b>
</h4>
<br><br><br>
<div class="text-center" style="font-size:16px"></div>
<h2>Part II: Attribute </h2>
<br>
<img src="img/attribute/8iVFBvox10YPSNR.png" style="width:30%;">  
<img src="img/attribute/AverageCTCsamples1112bitYPSNR.png" style="width:30%;">  
<img src="img/attribute/Fordq1mmPSNR.png" style="width:30%;">  
<h4> <b>R-D comparison for static attribute coding. </b>
</h4>
<br>
<img src="img/attribute/doubleLossy/longdress_PCQM.png" style="width:30%;">  
<img src="img/attribute/doubleLossy/loot_PCQM.png" style="width:30%;">  
<img src="img/attribute/doubleLossy/solider_PCQM.png" style="width:30%;">  
<h4> <b>R-D comparison for lossy compression of geometry & attribute. </b>
</h4>
<br>
<img src="img/attribute/dynamic/basketball_player_Y-PSNR.png" style="width:30%;">  
<img src="img/attribute/dynamic/soldier_Y-PSNR.png" style="width:30%;">  
<img src="img/attribute/dynamic/Ford_q1mm_PSNR.png" style="width:30%;">  
<h4> <b>R-D comparison for dynamic attribute coding. </b>
</h4>
<br>
<img src="img/attribute/vis_longdress_lossyA.png" style="width:80%;">  
<h4> <b>Qualitative results of reconstructed point clouds
with lossy attribute coding. </b>
</h4>
<br>
<img src="img/attribute/vis_longdress_lossyAlossyG.png" style="width:80%;">  
<h4> <b>Qualitative visualization of reconstructed point clouds
with lossy compression mode of both geometry & attribute. </b>
</h4>
</div>
</div>
</div>
</section>
<!-- <section id="introduction" class="container services">
<div class="row">
<div class="col-sm-4">
<h2>Component</h2>
<h4>Unicorn can compress the geometry and attribute information, either separately or jointly, of an input point cloud.</h4>
</div>
<div class="col-sm-4">
<h2>Tool</h2>
<h4>Unicorn flexibly supports the static and dynamic coding of point clouds in either lossless or lossy mode.</h4>
</div>
<div class="col-sm-4">
<h2>Source</h2>
<h4>Unicorn demonstrates the leading performance for diverse types, including solid, dense, and sparse object point clouds, as well as scant LiDAR.</h4>
</div>
</div>
</section> -->
<!-- <section class="container features">
<div class="row">
<div class="col-lg-12 text-center">
<div class="navy-line"></div>
<h1>Contributions of our Unicorn</h1>
</div>
</div>
<div class="row features-block">
<div class="col-sm-1">
</div>
<div class="col-sm-5">
<br/>
<br/>
<h2 style="color:rgb(53,53,53);"><b>Comprehensive coding metric</b></h2>
<h4 style="color:rgb(88,88,88);">Unicorn is the first, versatile, learning-based solution.</h4>
<br/>
<br/>
<h2 style="color:rgb(53,53,53);"><b>Better compression performance </b></h2>
<h4 style="color:rgb(88,88,88);">Unicorn provides significant performance gains to existing approaches.</h4>
<br/>
<br/>
<h2 style="color:rgb(53,53,53);"><b>Low computation complexity</b></h2>
<h4 style="color:rgb(88,88,88);">Unicorn is a low-complexity approach with comparable runtime measures to the G-PCC codec and variable-rate coding capability using a single neural model. </h4>
</div>
<div class="col-sm-5">
<img src="img/framework.png" alt="dashboard" class="img-responsive pull-right">
</div>
<div class="col-sm-1">
</div>
</div>
</section> -->
<!-- <section id="codes" class="codes">
<div class="container">
<div class="row m-b-lg">
<div class="col-lg-12 text-center">
<div class="navy-line"></div>
<h1>Codes and papers</h1>
<h4>These are links to our papers and codes, please cite us when using them.</h4>
</div>
</div>
</div>
</div>
</section> -->
<section id="team" class="gray-section team">
<div class="container">
<div class="row m-b-lg">
<div class="col-lg-12 text-center">
<div class="navy-line"></div>
<h1>Our Team</h1>
<p>Team members contributed to Unicorn.</p>
</div>
</div>
<div class="row">
<div class="col-sm-4">
<div class="team-member">
<h4><span class="navy">Ma</span> Zhan</h4>
<p>Professor at Nanjing University. </p>
<p>Email: [email protected] </p>
</div>
</div>
<div class="col-sm-4">
<div class="team-member">
<h4><span class="navy">Ding</span> Dandan</h4>
<p>Associate Professor at Hangzhou Normal University.</p>
<p>Email: [email protected] </p>
</div>
</div>
<div class="col-sm-4">
<div class="team-member">
<h4><span class="navy">Chen</span> Tong</h4>
<p>Associate Researcher at Nanjing University.</p>
<p>Email: [email protected] </p>
</div>
</div>
<div class="col-sm-4">
<div class="team-member">
<h4><span class="navy">Wang</span> Jianqiang</h4>
<p>Ph.D. Candidate at Nanjing University.</p>
<p>Email: [email protected] </p>
</div>
</div>
<div class="col-sm-4">
<div class="team-member">
<h4><span class="navy">Xue</span> Ruixiang</h4>
<p>Ph.D. Candidate at Nanjing University.</p>
<p>Email: [email protected] </p>
</div>
</div>
<div class="col-sm-4">
<div class="team-member">
<h4><span class="navy">Li</span> Jiaxin</h4>
<p>Ph.D. Candidate at Nanjing University.</p>
<p>Email: [email protected] </p>
</div>
</div>
</div>
</div>
</section>
<!-- <section id="ref" class="container services">
<div class="col-lg-12 text-center">
<h1>References</h1>
</div>
<p>J. Wang, R. Xue, J. Li, D. Ding, Y. Lin and Z. Ma, "
A Versatile Point Cloud Compressor
Using Universal Multiscale Conditional Coding – Part I: Geometry,"
in IEEE Transactions on Pattern Analysis and Machine Intelligence,
doi: 10.1109/TPAMI.2024.3462938.
</p>
<p>J. Wang, R. Xue, J. Li, D. Ding, Y. Lin and Z. Ma, "
A Versatile Point Cloud Compressor
Using Universal Multiscale Conditional Coding – Part II: Attribute,"
in IEEE Transactions on Pattern Analysis and Machine Intelligence,
doi: 10.1109/TPAMI.2024.3462945.
</p>
</section> -->
<script src="js/jquery-2.1.1.js"></script>
<script src="js/pace.min.js"></script>
<script src="js/bootstrap.min.js"></script>
<script src="js/classie.js"></script>
<script src="js/cbpAnimatedHeader.js"></script>
<script src="js/wow.min.js"></script>
<script src="js/inspinia.js"></script>
</body>
</html>
<!-- <section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title">BibTeX</h2>
<pre><code>@ARTICLE{10682571,
author={Wang, Jianqiang and Xue, Ruixiang and Li, Jiaxin and Ding, Dandan and Lin, Yi and Ma, Zhan},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
title={A Versatile Point Cloud Compressor Using Universal Multiscale Conditional Coding – Part I: Geometry},
year={2024},
volume={},
number={},
pages={1-18},
doi={10.1109/TPAMI.2024.3462938}
}</code></pre>
<pre><code>@article{10682571,
author={Wang, Jianqiang and Xue, Ruixiang and Li, Jiaxin and Ding, Dandan and Lin, Yi and Ma, Zhan},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
title={A Versatile Point Cloud Compressor Using Universal Multiscale Conditional Coding – Part I: Geometry},
year={2024},
volume={},
number={},
pages={1-18},
doi={10.1109/TPAMI.2024.3462938}
}</code></pre>
</div>
</section> -->