-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
717 lines (697 loc) · 45.9 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
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
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<meta name="description"
content="We present a novel audio-visual Occlusion-Robust Gender Recognition and Age Estimation (ORAGEN) approach. The proposed approach is based on intermediate features of unimodal transformer-based models and two Multi-Task Cross-Modal Attention (MTCMA) blocks, which predict gender, age, and protective mask type using voice and facial characteristics.">
<meta name="keywords" content="ORAGEN, Multimodal Paralinguistics, Gender Recognition, Age Estimation, Protective Mask Detection, Deep Learning, Transformer">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Audio-Visual Occlusion-Robust Gender Recognition and Age Estimation Approach Based on Multi-Task Cross-Modal Attention</title>
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" rel="stylesheet">
<link rel="stylesheet" href="./static/css/bulma.min.css">
<link rel="stylesheet" href="./static/css/bulma-carousel.min.css">
<link rel="stylesheet" href="./static/css/bulma-slider.min.css">
<link rel="stylesheet" href="./static/css/index.css">
<link rel="icon" type="image/png" href="https://hci.nw.ru/img/favicon/favicon-96x96.png" sizes="96x96">
<link rel="icon" type="image/png" href="https://hci.nw.ru/img/favicon/favicon-196x196.png" sizes="196x196">
<link rel="icon" type="image/png" href="https://hci.nw.ru/img/favicon/favicon-228x228.png" sizes="228x228">
<link rel="shortcut icon" href="https://hci.nw.ru/img/favicon/favicon.ico" type="image/x-icon">
<link rel="apple-touch-icon" href="https://hci.nw.ru/img/favicon/apple-touch-icon-57x57.png" sizes="52x57">
<link rel="apple-touch-icon" href="https://hci.nw.ru/img/favicon/apple-touch-icon-60x60.png" sizes="60x60">
<link rel="apple-touch-icon" href="https://hci.nw.ru/img/favicon/apple-touch-icon-72x72.png" sizes="72x72">
<link rel="apple-touch-icon" href="https://hci.nw.ru/img/favicon/apple-touch-icon-76x76.png" sizes="76x76">
<link rel="apple-touch-icon" href="https://hci.nw.ru/img/favicon/apple-touch-icon-114x114.png" sizes="114x114">
<link rel="apple-touch-icon" href="https://hci.nw.ru/img/favicon/apple-touch-icon-120x120.png" sizes="120x120">
<link rel="apple-touch-icon" href="https://hci.nw.ru/img/favicon/apple-touch-icon-144x144.png" sizes="144x144">
<link rel="apple-touch-icon" href="https://hci.nw.ru/img/favicon/apple-touch-icon-152x152.png" sizes="152x152">
<link rel="apple-touch-icon" href="https://hci.nw.ru/img/favicon/apple-touch-icon-180x180.png" sizes="180x180">
<link href="https://fonts.googleapis.com/css?family=Merriweather:400,900,900i" rel="stylesheet">
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script src="./static/js/bulma-carousel.min.js"></script>
<script src="./static/js/bulma-slider.min.js"></script>
<script src="./static/js/comp-slider.js" defer></script>
<script defer src="./static/js/fontawesome.all.min.js"></script>
</head>
<body>
<a id="btt-button">
<svg xmlns="http://www.w3.org/2000/svg" width="30" height="30" viewBox="0 0 24 24"> <polygon id="btt-arr" fill="#BBB" points="12 6.586 3.293 15.293 4.707 16.707 12 9.414 19.293 16.707 20.707 15.293 12 6.586"/>
</svg>
</a>
<section class="hero">
<div class="hero-body sepline">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<h1 class="title is-1 publication-title">
Audio-Visual Occlusion-Robust Gender Recognition and Age Estimation Approach Based on Multi-Task Cross-Modal Attention
</h1>
<div class="is-size-5 publication-authors">
<span class="author-block">
<a href="https://hci.nw.ru/en/employees/10" target="_blank">Maxim Markitantov</a><sup>1,2</sup>,
</span>
<span class="author-block">
<a href="https://hci.nw.ru/en/employees/14" target="_blank">Elena Ryumina</a><sup>1,2</sup>,
</span>
<span class="author-block">
<a href="https://hci.nw.ru/en/employees/1" target="_blank">Alexey Karpov</a><sup>1,3</sup>,
</span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block">
<sup>1</sup> St. Petersburg Institute for Informatics and Automation of the Russian Academy of Sciences, <a href="https://spcras.ru/en/" target="_blank">St. Petersburg Federal Research Center of the Russian Academy of Sciences (SPC RAS)</a>, St. Petersburg, Russia
</span>
<br />
<span class="author-block">
<sup>2</sup> <a href="https://www.uni-ulm.de/en/" target="_blank">ULM University</a>, Ulm, Germany
</span>
<br />
<span class="author-block">
<sup>3</sup> <a href="https://en.itmo.ru/" target="_blank">ITMO University</a>, St. Petersburg, Russia
</span>
<br />
</div>
<div class="column has-text-centered">
<div class="publication-links">
<!-- PDF Link. -->
<!-- Code Link. -->
<span class="link-block">
<a href="https://github.com/markitantov/ORAGEN" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code</span>
</a>
</span>
<!-- Models Link. -->
<span class="link-block">
<a href="#" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Models (coming soon)</span>
</a>
</span>
<!-- Dataset Link. -->
<span class="link-block">
<a href="https://hci.nw.ru/en/pages/brave-masks" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="far fa-images"></i>
</span>
<span>Data</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="section sepline">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="content">
<h3 class="title is-3">Abstract</h3>
<p class="has-text-justified">
Gender recognition and age estimation are essential tasks within soft biometric systems, where identifying these characteristics supports a wide range of applications. In real-world scenarios, challenges such as partial facial occlusion complicate these tasks by obscuring crucial voice and facial characteristics. These challenges highlight the importance of development of robust and efficient approaches for gender recognition and age estimation. In this study, we develop a novel audio-visual Occlusion-Robust Gender Recognition and Age Estimation (ORAGEN) approach. The proposed approach is based on intermediate features of unimodal transformer-based models and two Multi-Task Cross-Modal Attention (MTCMA) blocks, which predict gender, age, and protective mask type using voice and facial characteristics. We conduct detailed cross-corpus experiments on the TIMIT, aGender, CommonVoice, LAGENDA, IMDB-Clean, AFEW, VoxCeleb2, and BRAVE-MASKS corpora. The proposed unimodal models outperform State-of-the-Art approaches for gender recognition and age estimation. We investigate the impact of various protective mask types on the performance of audio-visual gender recognition and age estimation. The results show that the current large-scale data are still insufficient for a robust gender recognition and age estimation in partial facial occlusion conditions. On the Test subset of the VoxCeleb2 corpus, the proposed approach showed Unweighted Average Recall (UAR) of 99.51%, Mean Absolute Error (MAE) of 5.42, and UAR of 100% for gender recognition, age estimation, and protective mask type recognition, respectively, while on the Test subset of the BRAVE-MASKS corpus, it showed UAR=96.63%, MAE=7.52, and UAR=95.87%, for the same tasks. These results indicate that using data of people wearing protective masks, as well as including the protective mask type recognition task, yields performance gains on all tasks considered. ORAGEN can be integrated into the OCEAN-AI framework for optimizing HR processes, as well as into expert systems with practical applications in various domains including forensics, healthcare, and industrial safety.
</p>
</div>
</div>
</div>
</div>
</section>
<section class="section sepline">
<div class="container is-widescreen">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="content">
<h3 class="title is-3">Pipeline of the Proposed Approach</h3>
<img class="img-method" src="./static/img/proposed_approach.svg" alt="Proposed approach">
</div>
<div class="content has-text-justified">
<p>
Transformer layer and transformer block have different architecture. FCL refers to Fully-Connected Layer, SDPSA to Scaled Dot-Product Self-Attention, MTCMA to Multi-Task Cross-Modal Attention, VAD to Voice Activity Detection, NM to “No mask” class, TM to “Tissue mask”, MM to “Medical Mask”, PM to “Protective mask (FFP2/FFP3)”, PFS to “Protective face shield”, and R to “Respirator”
</p>
</div>
</div>
</div>
</div>
</section>
<section class="section sepline">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="content has-text-centered">
<h3 class="title is-3">Architecture of the Proposed Model for Audio-Visual Multi-Task Fusion</h3>
<img class="img-method" src="./static/img/audio_visual_model.svg" alt="Audio-visual model">
</div>
<div class="content has-text-justified">
<p>
MTCMA refers to Multi-Task Cross-Modal Attention. MLP to Multi Layer Perceptron. GELU to Gaussian Error Linear Unit.
</p>
</div>
</div>
</div>
</div>
</section>
<section class="section sepline">
<div class="container is-widescreen">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="content has-text-centered">
<h3 class="title is-3">Research Data</h3>
<table class="table">
<thead>
<tr>
<th class="has-text-left">Corpus</th>
<th class="has-text-left">Modalities</th>
<th class="has-text-left">Language</th>
<th class="has-text-left">Speaker Distribution</th>
<th class="has-text-left">Audio/Images Distribution</th>
<th class="has-text-left">File Format</th>
<th class="has-text-left">Annotation</th>
</tr>
</thead>
<tbody>
<tr>
<td>TIMIT</td>
<td>A</td>
<td>English</td>
<td class="has-text-right">192 females, 438 males, Aged 20-75 y.o., Mean is 29.9 y.o., Std. dev. is 7.9 y.o.</td>
<td>Total dur.: 6 hours, Num. of utt.: 6300, Avg. utt. dur.: 3.1 s.</td>
<td>PCM WAV, 16 kHz, Mono/16-bits</td>
<td>Gender, Age</td>
</tr>
<tr>
<td>aGender*</td>
<td>A</td>
<td>German</td>
<td class="has-text-right">335 females, 329 males, 106 children, Aged 7-80 y.o., Mean is 39.3 y.o., Std. dev. is 21.4 y.o.</td>
<td>Total dur.: 38.1 hours, Num. of utt.: 53074, Avg. utt. dur.: 2.5 s.</td>
<td>PCM WAV, 8 kHz, Mono/8-bits</td>
<td>Gender, Age</td>
</tr>
<tr>
<td>CommonVoice</td>
<td>A</td>
<td>English, German, Russian</td>
<td class="has-text-right">4352 females, 16828 males, Aged 13-99 y.o.</td>
<td>Total dur.: 5203 hours, Num. of utt.: 3583K, Avg. utt. dur.: 5.2 s.</td>
<td>MPEG, 48 kHz</td>
<td>Gender, Age group</td>
</tr>
<tr>
<td>LAGENDA</td>
<td>V</td>
<td>N/A</td>
<td class="has-text-right">42735 females, 41457 males, Aged 0-95 y.o., Mean is 36.8 y.o., Std. dev. is 21.6 y.o.</td>
<td class="has-text-right">67K images</td>
<td>JPG, Various resolutions</td>
<td>Gender, Age</td>
</tr>
<tr>
<td>IMDB-Clean</td>
<td>V</td>
<td>N/A</td>
<td class="has-text-right">127K females, 158K males, Aged 1-95 y.o., Mean is 37.1 y.o., Std. dev. is 12.8 y.o.</td>
<td class="has-text-right">296K images</td>
<td>JPG, Various resolutions</td>
<td>Gender, Age</td>
</tr>
<tr>
<td>AFEW*</td>
<td>A, V</td>
<td>English</td>
<td class="has-text-right">451 females, 705 males, Aged 5-76 y.o., Mean is 35.2 y.o., Std. dev. is 13.4 y.o.</td>
<td>Total dur.: 0.8 hours, Num. of utt.: 1156, Avg. utt. dur.: 2.5 s.</td>
<td>MPEG, 16 kHz, MPEG-4, 720x568, 25 FPS</td>
<td>Gender, Age</td>
</tr>
<tr>
<td>VoxCeleb2</td>
<td>A, V</td>
<td>English</td>
<td class="has-text-right">5333 females, 8888 males, Aged 10-95 y.o., Mean is 40.8 y.o., Std. dev. is 14.1 y.o.</td>
<td>Total dur.: 238 hours, Num. of utt.: 109K, Avg. utt. dur.: 7.8 s.</td>
<td>AAC, 16 kHz, MPEG-4, 224x224, 25 FPS</td>
<td>Gender, Age</td>
</tr>
<tr>
<td>BRAVE-MASKS</td>
<td>A, V</td>
<td>Russian</td>
<td class="has-text-right">15 females, 15 males, Aged 19-86 y.o., Mean is 40.8 y.o., Std. dev. is 19.0 y.o.</td>
<td>Total dur.: 21 hours, Num. of utt.: 14940, Avg. utt. dur.: 5.0 s.</td>
<td>PCM WAV, 48 kHz, Mono/16-bits, MPEG-4, 3840x2160, 30/60 FPS</td>
<td>Protective mask type, Gender, Age</td>
</tr>
</tbody>
</table>
</div>
<div class="content has-text-justified">
<p>
Metadata statistics of the corpora used. Std. dev. refers to standard deviation, avg. to average, utt. to utterance, dur. to duration, N/A to not available. * means that distribution of the Test subset is not available.
</p>
</div>
</div>
</div>
</div>
</section>
<section class="section sepline">
<div class="container is-full-screen">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="content has-text-centered" style="position: relative;">
<h3 class="title is-3">Exeperimental Results</h3>
<div class="slider">
<div class="slider__line">
<div class="slider__item">
<table class="table">
<thead>
<tr class="has-text-right">
<th class="has-text-left">Model</th>
<th class="has-text-left">Train subset</th>
<th>TIMIT</th>
<th>aGender</th>
<th>CommonVoice</th>
<th>VoxCeleb2</th>
<th colspan="2">BRAVE-MASKS</th>
</tr>
<tr class="has-text-right">
<th colspan="6"></th>
<th>Dev.</th>
<th>Test</th>
</tr>
</thead>
<tbody>
<tr>
<th colspan="8" class="centered">Gender recognition (UAR, %)</th>
</tr>
<tr>
<td>Burkhardt et al. (2023)</td>
<td>All corpora</td>
<td class="has-text-right">98.60</td>
<td class="has-text-right">86.15</td>
<td class="has-text-right">92.20</td>
<td class="has-text-right">89.50</td>
<td class="has-text-right">84.16</td>
<td class="has-text-right">85.15</td>
</tr>
<tr>
<td><b>W2V2-based</b></td>
<td><b>All corpora</b></td>
<td class="has-text-right"><b>98.60</b></td>
<td class="has-text-right"><b>87.17</b></td>
<td class="has-text-right"><b>92.59</b></td>
<td class="has-text-right"><b>90.00</b></td>
<td class="has-text-right"><b>85.14</b></td>
<td class="has-text-right"><b>86.22</b></td>
</tr>
<tr>
<td>HuBERT-based</td>
<td>All corpora</td>
<td class="has-text-right">97.97</td>
<td class="has-text-right">86.99</td>
<td class="has-text-right">92.22</td>
<td class="has-text-right">89.83</td>
<td class="has-text-right">84.95</td>
<td class="has-text-right">86.21</td>
</tr>
<tr>
<th colspan="8" class="centered">Age estimation (MAE)</th>
</tr>
<tr>
<td>Burkhardt et al. (2023)</td>
<td>All corpora</td>
<td class="has-text-right">7.10</td>
<td class="has-text-right">10.80</td>
<td class="has-text-right">11.15</td>
<td class="has-text-right">10.31</td>
<td class="has-text-right">13.96</td>
<td class="has-text-right">13.09</td>
</tr>
<tr>
<td><b>W2V2-based</b></td>
<td><b>All corpora</b></td>
<td class="has-text-right"><b>6.90</b></td>
<td class="has-text-right"><b>10.60</b></td>
<td class="has-text-right"><b>10.47</b></td>
<td class="has-text-right"><b>9.91</b></td>
<td class="has-text-right"><b>11.65</b></td>
<td class="has-text-right"><b>11.89</b></td>
</tr>
<tr>
<td>HuBERT-based</td>
<td>All corpora</td>
<td class="has-text-right">7.00</td>
<td class="has-text-right">10.92</td>
<td class="has-text-right">11.32</td>
<td class="has-text-right">10.34</td>
<td class="has-text-right">13.01</td>
<td class="has-text-right">12.23</td>
</tr>
</tbody>
</table>
<div class="content has-text-justified">
<p>
Audio-based experimental results. "All corpora" includes only the TIMIT, aGender, and CommonVoice corpora. Dev. refers to development.
</p>
</div>
</div>
<div class="slider__item">
<table class="table">
<thead>
<tr class="has-text-right">
<th class="has-text-left">Model</th>
<th class="has-text-left">Train subset</th>
<th>LAGENDA</th>
<th>AFEW</th>
<th colspan="2">IMDB-Clean</th>
<th>VoxCeleb2</th>
<th colspan="2">BRAVE-MASKS</th>
</tr>
<tr class="has-text-right">
<th colspan="4"></th>
<th>Dev.</th>
<th>Test</th>
<th></th>
<th>Dev.</th>
<th>Test</th>
</tr>
</thead>
<tbody>
<tr>
<th colspan="9" class="centered">Gender recognition (UAR, %)</th>
</tr>
<tr>
<td>Kuprashevich et al. (2023)</td>
<td>IMDB-Clean</td>
<td class="has-text-right">91.11</td>
<td class="has-text-right">94.60</td>
<td class="has-text-right"><b>99.70</b></td>
<td class="has-text-right"><b>99.40</b></td>
<td class="has-text-right">98.35</td>
<td class="has-text-right">87.84</td>
<td class="has-text-right">93.39</td>
</tr>
<tr>
<td>SDPSA-based</td>
<td>All corpora</td>
<td class="has-text-right"><b>92.89</b></td>
<td class="has-text-right">95.16</td>
<td class="has-text-right">98.49</td>
<td class="has-text-right">98.37</td>
<td class="has-text-right">98.37</td>
<td class="has-text-right">88.12</td>
<td class="has-text-right"><b>94.44</b></td>
</tr>
<tr>
<td>GSA-based</td>
<td>All corpora</td>
<td class="has-text-right">92.72</td>
<td class="has-text-right"><b>95.41</b></td>
<td class="has-text-right">98.70</td>
<td class="has-text-right">98.46</td>
<td class="has-text-right">98.25</td>
<td class="has-text-right"><b>89.83</b></td>
<td class="has-text-right">90.16</td>
</tr>
<tr>
<th colspan="9" class="centered">Age estimation (MAE)</th>
</tr>
<tr>
<td>Kuprashevich et al. (2023)</td>
<td>IMDB-Clean</td>
<td class="has-text-right">5.40</td>
<td class="has-text-right">6.09</td>
<td class="has-text-right"><b>3.48</b></td>
<td class="has-text-right"><b>4.28</b></td>
<td class="has-text-right">7.32</td>
<td class="has-text-right">9.60</td>
<td class="has-text-right">9.22</td>
</tr>
<tr>
<td>SDPSA-based</td>
<td>All corpora</td>
<td class="has-text-right">5.18</td>
<td class="has-text-right">5.62</td>
<td class="has-text-right">5.23</td>
<td class="has-text-right">5.47</td>
<td class="has-text-right"><b>5.97</b></td>
<td class="has-text-right"><b>8.86</b></td>
<td class="has-text-right"><b>8.71</b></td>
</tr>
<tr>
<td>GSA-based</td>
<td>All corpora</td>
<td class="has-text-right">5.05</td>
<td class="has-text-right"><b>5.59</b></td>
<td class="has-text-right">5.03</td>
<td class="has-text-right">5.43</td>
<td class="has-text-right">6.59</td>
<td class="has-text-right">9.65</td>
<td class="has-text-right">9.42</td>
</tr>
</tbody>
</table>
<div class="content has-text-justified">
<p>
Video-based experimental results. "All corpora" includes only LAGENDA, AFEW, and IMDB-Clean corpora. Dev. refers to development.
</p>
</div>
</div>
</div>
</div>
<div class="slider__btn-prev">
<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none">
<path d="M3 12L21 24V0L3 12Z" fill="#bbb"/>
</svg>
</div>
<div class="slider__btn-next">
<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24">
<path d="M21 12l-18 12v-24z" fill="#bbb"/>
</svg>
</div>
<div class="slider__wrapper">
<div class="slider__dot active-dot"></div>
<div class="slider__dot"></div>
</div>
</div>
<div class="content has-text-centered">
<table class="table">
<thead>
<tr class="has-text-right">
<th class="has-text-left">Method</th>
<th>VoxCeleb2</th>
<th>BRAVE-MASKS</th>
</tr>
</thead>
<tbody>
<tr>
<th colspan="3" class="centered">Gender recognition (UAR, %)</th>
</tr>
<tr>
<td>Audio-based</td>
<td class="has-text-right">90.00</td>
<td class="has-text-right">86.22</td>
</tr>
<tr>
<td>Video-based</td>
<td class="has-text-right">98.37</td>
<td class="has-text-right">94.44</td>
</tr>
<tr>
<td>Early fusion</td>
<td class="has-text-right">98.90</td>
<td class="has-text-right">94.80</td>
</tr>
<tr>
<td>Intermediate fusion</td>
<td class="has-text-right">99.11</td>
<td class="has-text-right">94.95</td>
</tr>
<tr>
<td>Late fusion</td>
<td class="has-text-right">99.02</td>
<td class="has-text-right">97.21</td>
</tr>
<tr>
<th colspan="3" class="centered">Age estimation (MAE)</th>
</tr>
<tr>
<td>Audio-based</td>
<td class="has-text-right">9.91</td>
<td class="has-text-right">11.89</td>
</tr>
<tr>
<td>Video-based</td>
<td class="has-text-right">5.97</td>
<td class="has-text-right">8.71</td>
</tr>
<tr>
<td>Early fusion</td>
<td class="has-text-right">5.80</td>
<td class="has-text-right">9.00</td>
</tr>
<tr>
<td>Intermediate fusion</td>
<td class="has-text-right">5.68</td>
<td class="has-text-right">8.73</td>
</tr>
<tr>
<td>Late fusion</td>
<td class="has-text-right">6.51</td>
<td class="has-text-right">7.23</td>
</tr>
</tbody>
</table>
</div>
<div class="content has-text-justified">
<p>
Experimental results on audio-visual multi-task gender recognition (classification task, 2 classes) and age estimation (regression task) on the Test subsets of the VoxCeleb2 and BRAVE-MASKS corpora.
</p>
</div>
</div>
</div>
</div>
</section>
<section class="section sepline">
<div class="container is-widescreen">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="content has-text-centered">
<h3 class="title is-3">Multi-Task Cross-Modal Attention Visualization</h3>
<img class="img-method" src="./static/img/heatmaps.svg" alt="Cross-modal attention visualization">
</div>
</div>
</div>
</div>
</section>
<section class="section sepline">
<div class="container is-widescreen">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="content has-text-centered">
<h3 class="title is-3">Protective Mask Type Influence on Gender Recognition and Age Estimation Performance</h3>
<table class="table">
<thead>
<tr class="has-text-right">
<th class="has-text-left">Data</th>
<th>All masks</th>
<th>No mask</th>
<th>Tissue mask</th>
<th>Medical mask</th>
<th>Protective mask</th>
<th>Respirator (FFP2/FFP3)</th>
<th>Protective face shield</th>
</tr>
</thead>
<tbody>
<tr>
<th colspan="8" class="centered">Gender recognition (UAR, %)</th>
</tr>
<tr>
<td>Audio-visual</td>
<td class="has-text-right">94.95</td>
<td class="has-text-right">98.33</td>
<td class="has-text-right">96.38</td>
<td class="has-text-right">97.69</td>
<td class="has-text-right">98.02</td>
<td class="has-text-right">90.56</td>
<td class="has-text-right">89.00</td>
</tr>
<tr>
<td>Audio-only</td>
<td class="has-text-right">93.98</td>
<td class="has-text-right">98.33</td>
<td class="has-text-right">96.77</td>
<td class="has-text-right">94.75</td>
<td class="has-text-right">97.80</td>
<td class="has-text-right">88.61</td>
<td class="has-text-right">89.00</td>
</tr>
<tr>
<td>Video-only</td>
<td class="has-text-right">93.06</td>
<td class="has-text-right">97.67</td>
<td class="has-text-right">96.11</td>
<td class="has-text-right">97.69</td>
<td class="has-text-right">94.74</td>
<td class="has-text-right">89.91</td>
<td class="has-text-right">89.00</td>
</tr>
<tr>
<th colspan="8" class="centered">Age estimation (MAE)</th>
</tr>
<tr>
<td>Audio-visual</td>
<td class="has-text-right">8.73</td>
<td class="has-text-right">4.83</td>
<td class="has-text-right">7.74</td>
<td class="has-text-right">11.12</td>
<td class="has-text-right">10.44</td>
<td class="has-text-right">15.27</td>
<td class="has-text-right">5.63</td>
</tr>
<tr>
<td>Audio-only</td>
<td class="has-text-right">11.91</td>
<td class="has-text-right">10.93</td>
<td class="has-text-right">12.09</td>
<td class="has-text-right">15.39</td>
<td class="has-text-right">12.04</td>
<td class="has-text-right">17.44</td>
<td class="has-text-right">8.85</td>
</tr>
<tr>
<td>Video-only</td>
<td class="has-text-right">8.98</td>
<td class="has-text-right">5.92</td>
<td class="has-text-right">8.78</td>
<td class="has-text-right">12.29</td>
<td class="has-text-right">11.06</td>
<td class="has-text-right">19.87</td>
<td class="has-text-right">6.62</td>
</tr>
</tbody>
</table>
</div>
<div class="content has-text-justified">
<p>
Influence of particular protective mask type on gender recognition (classification task, 2 classes) and age estimation (regression task) performance on the Test subset of the BRAVE-MASKS corpus.
</p>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-widescreen">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="content has-text-centered">
<h3 class="title is-3">Possible Application of the Proposed Approach for Optimizing HR Processes</h3>
<img class="img-method" src="./static/img/application.svg" alt="Possible application">
</div>
</div>
</div>
</div>
</section>
<footer class="footer" style="background-color: rgb(247, 247, 247);">
<div class="container">
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p>
This page was built using the <a href="https://github.com/SMIL-SPCRAS/ORAGEN" target="_blank">ORAGEN page repository</a>, which was adopted from the <a href="https://nerfies.github.io" target="_blank">Nerfies</a> project page.
You are free to borrow the of this website, we just ask that you link back to this page in the footer. <br> This website is licensed under a <a rel="license" href="http://creativecommons.org/licenses/by-sa/4.0/" target="_blank">Creative
Commons Attribution-ShareAlike 4.0 International License</a>.
</p>
</div>
</div>
</div>
</div>
</footer>
</body>
</html>