-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathindex.html
482 lines (447 loc) · 19.9 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
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<meta name="description"
content="Towards Practical Single-shot Motion Synthesis.">
<meta name="keywords" content="Moverse, AI, Motion Synthesis, Single-shot Synthesis, Single-sample Synthesis">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Towards Practical Single-shot Motion Synthesis</title>
<!-- Global site tag (gtag.js) - Google Analytics
<script async src="https://www.googletagmanager.com/gtag/js?id=G-XXXXXXXXXXX"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag() {
dataLayer.push(arguments);
}
gtag('js', new Date());
gtag('config', 'G-XXXXXXXXXXXXX');
</script> -->
<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/fontawesome.all.min.css">
<link rel="stylesheet"
href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="./static/css/index.css">
<link rel="icon" href="./static/images/favicon.png">
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script defer src="./static/js/fontawesome.all.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/index.js"></script>
<link rel="stylesheet" href="./static/css/dics.css">
<script src="./static/js/dics.js"></script>
<link rel="stylesheet" href="./static/css/BeerSlider.css"/>
<script src="./static/js/BeerSlider.js"></script>
<script id="MathJax-script" async src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js"></script>
</head>
<body>
<nav class="navbar" role="navigation" aria-label="main navigation">
<div class="navbar-brand">
<a role="button" class="navbar-burger" aria-label="menu" aria-expanded="false">
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
</a>
</div>
<div class="navbar-menu">
<div class="navbar-start" style="flex-grow: 1; justify-content: center;">
<a class="navbar-item" href="https://www.github.com/moverseai">
<span class="icon">
<i class="fas fa-home"></i>
</span>
</a>
<div class="navbar-item has-dropdown is-hoverable">
<a class="navbar-link">
More Research
</a>
<div class="navbar-dropdown">
<a class="navbar-item" href="https://moverseai.github.io/noise-tail">
Long-tail
</a>
<a class="navbar-item" href="https://moverseai.github.io/bundle">
BundleMoCap
</a>
<a class="navbar-item" href="https://moverseai.github.io/single-shot">
Single-shot
</a>
</div>
</div>
</div>
</div>
</nav>
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<img width=8% src="./static/images/mov_icon.png" class="center">
<br><br>
<!-- <br> -->
<h1 class="title is-1 publication-title">Towards Practical Single-shot Motion Synthesis</h1>
<div class="is-size-5 publication-authors">
<span class="author-block">
<a href="https://www.linkedin.com/in/kroditakis/">Konstantinos Roditakis</a>, </span>
<span class="author-block">
<a href="https://spthermo.github.io/">Spyridon Thermos</a>, </span>
<span class="author-block">
<a href="https://zokin.github.io">Nikolaos Zioulis</a></span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block"><a href="https://www.moverse.ai">Moverse</a></span>
<h2>CVPR 2024, AI for 3D Generation Workshop</h2>
</div>
<div class="column has-text-centered">
<div class="publication-links">
<span class="link-block">
<a href="https://arxiv.org/pdf/2406.01136"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>arXiv</span>
</a>
</span>
<span class="link-block">
<a href="./static/misc/poster.pdf"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Poster</span>
</a>
</span>
<span class="link-block">
<a href="https://github.com/moverseai/ganimateer"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code (Coming Soon)</span>
</a>
</span>
</div>
<br>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- Teaser. -->
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column is-centered has-text-centered">
<div class="container has-text-centered is-max-desktop">
<img width=100% src="./static/images/teaser.png" class="center">
</div>
</div>
</div>
</div>
</section>
<!--/ Teaser. -->
<!-- Abstract. -->
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-2">Abstract</h2>
<div class="content has-text-justified">
<p>
Despite the recent advances in the so-called "cold start" generation from text prompts, their needs in data and computing resources, as well as the ambiguities around intellectual property and privacy concerns pose certain counterarguments for their utility.
An interesting and relatively unexplored alternative has been the introduction of unconditional synthesis from a single sample, which has led to interesting generative applications.
</p>
<p>
In this paper we focus on single-shot motion generation and more specifically on accelerating the training time of a Generative Adversarial Network (GAN).
In particular, we tackle the challenge of GAN's equilibrium collapse when using mini-batch training by carefully annealing the weights of the loss functions that prevent mode collapse.
Additionally, we perform statistical analysis in the generator and discriminator models to identify correlations between training stages and enable transfer learning.
Our improved GAN achieves competitive quality and diversity on the Mixamo benchmark when compared to the original GAN architecture and a single-shot diffusion model, while being up to \(\times 6.8\) faster in training time from the former and \(\times 1.75\) from the latter.
</p>
<p>
Finally, we demonstrate the ability of our improved GAN to mix and compose motion with a single forward pass.
</p>
</div>
</div>
</div>
</section>
<!--/ Abstract-->
<section class="section">
<div class="container is-max-desktop">
<!-- columns is-full-width -->
<div class="columns is-centered">
<div class="column is-centered has-text-centered">
<h2 class="title is-2">Mixamo-based Generation</h2>
<p>
We provide some indicative applications of our single-shot GAN that <b>do not need</b> any re-training.
</p>
<br>
<br>
<h3 class="title is-3">Single-Motion Variations</h3>
<p>
We use our single-shot GAN trained on the "breakdance freezes" sequence from the <a href="https://www.mixamo.com/">Mixamo</a> dataset to generate variations of the "breakdance freezes" sample by sampling different codes from a Gaussian distribution.
For visualization purposes we use the "<i>Michelle</i>" character provided by <a href="https://www.mixamo.com/">Mixamo</a>.
</p>
<video muted="" loop="" playsinline="" controls="" width="100%">
<source src="./static/videos/mixamo_breakdance_freezes_gen.mp4" type="video/mp4">
</video>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<!-- columns is-full-width -->
<div class="columns is-centered">
<div class="column is-centered has-text-centered">
<h3 class="title is-3">Upper-body Motion Composition</h3>
<p>
We use the Mixamo motion "swing dancing" as an example input sequence, keeping the lower-body unaltered (fixed) and generating 7 alternative - but natural - versions of the upper-body.
The displayed result is rendering using the "<i>Jackie</i>" character provided by <a href="https://www.mixamo.com/">Mixamo</a>.
</p>
<video muted="" loop="" playsinline="" controls="" width="100%">
<source src="./static/videos/mixamo_swing_dancing_gen.mp4" type="video/mp4">
</video>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<!-- columns is-full-width -->
<div class="columns is-centered">
<div class="column is-centered has-text-centered">
<h3 class="title is-3">Lower-body Motion Composition</h3>
<p>
Changing to the "salsa dancing" input sequence, we now keep the upper-body fixed and sample partial codes for 7 alternative versions of the lower-body.
The "<i>Michelle</i>" character is used for visualization.
</p>
<video muted="" loop="" playsinline="" controls="" width="100%">
<source src="./static/videos/mixamo_salsa_dancing_gen.mp4" type="video/mp4">
</video>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<!-- columns is-full-width -->
<div class="columns is-centered">
<div class="column is-centered has-text-centered">
<h3 class="title is-3">Crowd Animation</h3>
<p>
Following <a href="https://peizhuoli.github.io/ganimator">[1]</a> and <a href="https://sinmdm.github.io/SinMDM-page">[2]</a>
we provide an example of a "crowd" dancing using generated variations of the "dancing" sequence of the <a href="https://www.mixamo.com/">Mixamo</a> dataset.
Again, for rendered result the "<i>Jackie</i>" character is used.
</p>
<video muted="" loop="" playsinline="" controls="" width="100%">
<source src="./static/videos/mixamo_dancing_crowd_gen.mp4" type="video/mp4">
</video>
</div>
</div>
</div>
</section>
<!-- References. -->
<section class="section">
<div class="container is-max-desktop content">
<h2 class="title is-3">References</h2>
<div class="content has-text-justified">
<p>
[1] <a href="https://peizhuoli.github.io/ganimator">GANimator</a>: Neural motion synthesis from a single sequence.
</p>
<p>
[2] <a href="https://sinmdm.github.io/SinMDM-page">SinMDM</a>: Single motion diffusion.
</p>
</div>
</div>
</section>
<!--/ References. -->
<section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title">BibTeX</h2>
<pre><code>@inproceedings{roditakis2024singleshot,
author = {Roditakis, Konstantinos, and Thermos, Spyridon, and Zioulis, Nikolaos},
title = {Towards Practical Single-Shot Motion Synthesis},
booktitle = {IEEE/CVF Computer Vision and Pattern Recognition (CVPR) AI for 3D Generation Workshop},
url = {https://moverseai.github.io/single-shot},
month = {June},
year = {2024}
}</code></pre>
</div>
</section>
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<!-- <img width=8% src="./static/images/mov_icon.png" class="center"> -->
<br><br>
<h1 class="title is-1 publication-title">Controlling Diversity in Single-shot Motion Synthesis</h1>
<div class="is-size-5 publication-authors">
<span class="author-block">
<a href="https://www.linkedin.com/in/eleni-tselepi-14b895245/?originalSubdomain=gr">Eleni Tselepi</a><sup>1, 2</sup>,</span>
<span class="author-block">
<a href="https://spthermo.github.io/">Spyridon Thermos</a><sup>1</sup>,</span>
<span class="author-block">
<a href="https://tzole1155.github.io/">Georgios Albanis</a><sup>1</sup>,</span>
<span class="author-block">
<a href="https://tofis.github.io/">Chatzitofis Anargyros</a><sup>1</sup>,</span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block"><a href="https://www.moverse.ai">Moverse</a></span>
<span class="author-block"><sup>2</sup><a href="https://www.e-ce.uth.gr/">University of Thessaly</a></span>
<h2>SIGGRAPH Asia 2024 (Poster)</h2>
</div>
<div class="column has-text-centered">
<div class="publication-links">
<!-- PDF Link. -->
<span class="link-block">
<a href="https://dl.acm.org/doi/pdf/10.1145/3681756.3697909"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="ai ai-acmdl"></i>
</span>
<span>Paper</span>
</a>
</span>
<span class="link-block">
<a href="https://drive.google.com/file/d/14u2u-toqvM7SCJPYlhddK4aKDy-fsmbU/view?usp=sharing"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Poster</span>
</a>
</span>
<!-- Code Link. -->
<!-- <span class="link-block">
<a href="https://github.com/moverseai/noise-tail"
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> -->
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- Teaser. -->
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column is-centered has-text-centered">
<div class="container has-text-centered is-max-desktop">
<img width=100% src="./static/images/teaser-asia.png" class="center">
</div>
</div>
</div>
</div>
</section>
<!--/ Teaser. -->
<!-- Abstract. -->
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-2">Abstract</h2>
<div class="content has-text-justified">
<p>
We consider the task of controllable and diverse motion synthesis from a single sequence as an alternative to the data-dependent text-to-motion methods,
which pose ambiguities in data ownership and privacy. Recent works in hierarchical single-shot synthesis have paved the path for unconditional generation and editing tools,
however the methods that focus on 3D animation have failed to control the diversity of the generated motions.
</p>
<p>
In this paper we propose the integration of the variational inference in single-shot GANs, aiming to encode and control the low-frequency generating factors of the single motion
sample. Our experiment showcases the ability of our VAE-GAN model to control the diversity of its generations, while preserving their plausibility and quality.
</p>
</div>
</div>
</div>
</section>
<!--/ Abstract-->
<!-- References. -->
<section class="section">
<div class="container is-max-desktop content">
<h2 class="title is-3">References</h2>
<div class="content has-text-justified">
<p>
[1] <a href="https://shirgur.github.io/hp-vae-gan/">Hierarchical Patch VAE-GAN:</a>: Generating Diverse Videos from a Single Sample.
</p>
<p>
[2] <a href="https://deepmotionediting.github.io/retargeting">Skeleton-Aware Networks</a> for Deep Motion Retargeting.
</p>
</div>
</div>
</section>
<!--/ References. -->
<section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title">BibTeX</h2>
<pre><code>@inproceedings{tselepi2024singleshot,
author = {Tselepi, Eleni, and Thermos, Spyridon, and Albanis, Georgios and Chatzitofis, Anargyros},
title = {Controlling Diversity in Single-shot Motion Synthesis},
booktitle = {SIGGRAPH Asia (Posters)},
url = {https://moverseai.github.io/single-shot},
month = {December},
year = {2024}
}</code></pre>
</div>
</section>
<section class="section" id="Acknowledgement">
<div class="container is-max-desktop content">
<h2 class="title">Acknowledgement</h2>
<p class="aligned-paragraph">
<img src="./static/images/eu_logo.png" class="left-image">
<img src="./static/images/emil.png" class="left-image">
<span class="right-text">Both projects have received financial support from the European Union’s Horizon Europe Research and Innovation Programme under Grant Agreement No 101070533, <a href="https://emil-xr.eu/"><i>EMIL-XR</i></a>.</span>
</p>
<!-- <div class="container has-text-centered is-max-desktop">
<img width=10% src="./static/images/emil.png" class="center">
This project has received funding from the European Union’s Horizon Europe Research and Innovation Programme under Grant Agreement No 101070533, <a href="https://emil-xr.eu/"><i>EMIL-XR</i></a>.
</div> -->
</div>
</div>
</section>
<footer class="footer">
<div class="container">
<div class="content has-text-centered">
<a class="icon-link"
href="https://arxiv.org/pdf/2406.01136">
<i class="fas fa-file-pdf"></i>
</a>
<a class="icon-link" href="https://github.com/moverseai" class="external-link" disabled>
<i class="fab fa-github"></i>
</a>
<a class="icon-link" href="https://discord.gg/bQc7B6qSPd" class="external-link" disabled>
<i class="fab fa-discord"></i>
</a>
</div>
<div class="columns is-centered">
<div class="content">
<p>
The website template has been adopted from <a href="https://nerfies.github.io" target="_blank">nerfies</a>.
</p>
</div>
</div>
</div>
</footer>
<script>
$.fn.BeerSlider = function ( options ) {
options = options || {};
return this.each(function() {
new BeerSlider(this, options);
});
};
$('.beer-slider').each( (function( index, el ) {
$(el).BeerSlider({start: $(el).data('beer-start')})
}));
</script>
</body>
</html>