-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathindex.html
More file actions
381 lines (360 loc) · 16.9 KB
/
index.html
File metadata and controls
381 lines (360 loc) · 16.9 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
<!DOCTYPE html>
<html lang="en">
<head>
<!-- Required meta tags -->
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Diffused Heads: Diffusion Models Beat GANs on Talking-Face Generation</title>
<!-- Compiled and minified CSS -->
<link
href="https://netdna.bootstrapcdn.com/twitter-bootstrap/2.3.2/css/bootstrap-combined.no-icons.min.css"
rel="stylesheet">
<link
href="https://netdna.bootstrapcdn.com/font-awesome/3.2.1/css/font-awesome.css"
rel="stylesheet">
<link rel="stylesheet"
href="https://cdnjs.cloudflare.com/ajax/libs/materialize/1.0.0/css/materialize.min.css">
<link
href="https://fonts.googleapis.com/css2?family=B612+Mono&family=Open+Sans&display=swap"
rel="stylesheet">
<link rel="stylesheet" href="css/style.css">
<link rel="stylesheet" href="css/additional.css">
<link rel="stylesheet" href="css/academicons.min.css" />
<link rel="apple-touch-icon" sizes="180x180" href="/diffusedheads/apple-touch-icon.png">
<link rel="icon" type="image/png" sizes="32x32" href="/diffusedheads/favicon-32x32.png">
<link rel="icon" type="image/png" sizes="16x16" href="/diffusedheads/favicon-16x16.png">
<link rel="manifest" href="/diffusedheads/site.webmanifest">
<link rel="mask-icon" href="/safari-pinned-tab.svg" color="#5bbad5">
<meta name="msapplication-TileColor" content="#da532c">
<meta name="theme-color" content="#ffffff">
</head>
<body>
<div class="container">
<div class="row">
<div class="col s12">
<h1 align="center">
<!-- ️<i class="fa fa-sliders" aria-hidden="true" /></i> -->
Diffused Heads: Diffusion Models Beat GANs <br>on Talking-Face Generation
<!-- <i class="fa fa-sliders" aria-hidden="true"></i> -->
</h1>
<!-- <h2 align="center">
Supplementary Material
</h2> -->
</div>
</div>
<!-- <div class="row">
<div class="col s1"></div>
<div class="col s2 center">
Michał Stypułkowski <sup>1</sup>
</div>
<div class="col s2 center">
Konstantinos Vougioukas <sup>2</sup>
</div>
<div class="col s2 center">
Sen He
</div>
</div> -->
<div class="row">
<div class="col s1"> </div>
<div class="col s12 center">
<p style="font-size:18px">
Michał Stypułkowski <sup>1</sup>
<span
style="margin-left: 1em ; margin-right: 1em;"> </span>
Konstantinos Vougioukas <sup>2</sup>
<span
style="margin-left: 1em ; margin-right: 1em;"> </span>
Sen He
<span
style="margin-left: 1em ; margin-right: 1em;"> </span>
Maciej Zięba <sup>3,4</sup>
<span
style="margin-left: 1em ; margin-right: 1em;"> </span>
Stavros Petridis <sup>2</sup>
<span
style="margin-left: 1em ; margin-right: 1em;"> </span>
Maja Pantic <sup>2</sup>
</p>
</div>
<div class="col s1"> </div>
</div>
<div class="row">
<div class="col s1"> </div>
<div class="col s10 center">
<sup>1</sup> University of Wrocław
<span
style="margin-left: 1em ; margin-right: 1em;"> </span>
<sup>2</sup> Imperial College London
<span
style="margin-left: 1em ; margin-right: 1em;"> </span>
<sup>3</sup> Wrocław University of Science and Technology
<span
style="margin-left: 1em ; margin-right: 1em;"> </span>
<sup>4</sup> Tooploox
</div>
<div class="col s1"> </div>
</div>
<div class="row">
</div>
<div class="row">
<div class="col s12 center">
<a class="btn-floating grey waves-effect waves-light tooltipped"
data-position="bottom"
data-tooltip="Code"
href="https://github.com/MStypulkowski/diffused-heads"
target="_blank">
<i class="fab fa-github"></i>
</a>
<a class="btn-floating grey waves-effect waves-light tooltipped"
data-position="bottom" data-tooltip="Abstract"
href="https://arxiv.org/abs/2301.03396"
target="_blank">
<i class="ai ai-arxiv"></i>
</a>
<a class="btn-floating grey waves-effect waves-light tooltipped"
data-position="bottom" data-tooltip="Paper"
href="diffused_heads.pdf"
target="_blank">
<i class="fas fa-file-pdf"></i> Paper
</a>
<!-- <a class="btn-floating grey waves-effect waves-light tooltipped"
data-position="bottom" data-tooltip="Dataset"
href="http://cvlab.ii.pw.edu.pl:9000/conerf-dataset/datasets.zip">
<i class="fas fa-database"></i> Paper
</a> -->
</div>
</div>
<hr>
<div class="row">
<div class="col s4 center">
<video style="width: 90%" loop muted autoplay>
<source
src="videos\1030_MTI_FEA_XX.mp4"
type="video/mp4">
</video>
</div>
<div class="col s4 center">
<video style="width: 90%" loop muted autoplay>
<source
src="videos\1081_DFA_HAP_XX.mp4"
type="video/mp4">
</video>
</div>
<div class="col s4 center">
<video style="width: 90%" loop muted autoplay>
<source
src="videos\1020_IOM_SAD_XX.mp4"
type="video/mp4">
</video>
</div>
</div>
<hr>
<div>
<div>
<p align="justify" style="font-size:18px">
Diffused Heads is the first method successfully using a diffusion model to generate talking faces.
Given a single identity frame and an audio clip containing speech, the model samples consecutive frames in an autoregressive manner,
preserving the identity, and modeling lip and head movement to match the audio input. Contrary to other methods, no additional guidance is required.
</p>
<!-- <div class="col s12 center"> -->
<!-- <video style="width: 100%; border: 1pt solid black;"
controls>
<source src="images\graphical_abstract.pdf" type="picture/pdf">
Teaser
</video> -->
<img src="images\graphical_abstract.png" alt="Overview", width=60%, class="center">
<!-- </div> -->
</div>
</div>
<hr>
<div>
<h2>Generated videos</h2>
<p align="justify" style="font-size:18px">
We present generated test samples from CREMA dataset, and generalization results. For more results and supplementary video, please contact us at <b>michal.stypulkowski[at]cs.uni.wroc.pl</b>.
</p>
<div id="combined-container" class="row">
<div class="col s12">
<div class="row">
<h3 style="font-variant: small-caps;">
CREMA
</h3>
<p align="justify" style="font-size:18px">
Test set examples from CREMA. None of the participants was included in the training set.
</p>
<div class="col s3 center">
<video style="width: 100%; border: 1pt solid black;"
controls>
<source src="videos\1081_TSI_HAP_XX.mp4" type="video/mp4">
</video>
</div>
<div class="col s3 center">
<video style="width: 100%; border: 1pt solid black;"
controls>
<source src="videos\1021_TAI_ANG_XX.mp4" type="video/mp4">
</video>
</div>
<div class="col s3 center">
<video style="width: 100%; border: 1pt solid black;"
controls>
<source src="videos\1089_IEO_ANG_LO.mp4" type="video/mp4">
</video>
</div>
<div class="col s3 center">
<video style="width: 100%; border: 1pt solid black;"
controls>
<source src="videos\1015_TIE_ANG_XX.mp4" type="video/mp4">
</video>
</div>
</div>
<div class="row">
<div class="col s3 center">
<video style="width: 100%; border: 1pt solid black;"
controls>
<source src="videos\1033_IOM_NEU_XX.mp4" type="video/mp4">
</video>
</div>
<div class="col s3 center">
<video style="width: 100%; border: 1pt solid black;"
controls>
<source src="videos\1052_IWW_ANG_XX.mp4" type="video/mp4">
</video>
</div>
<div class="col s3 center">
<video style="width: 100%; border: 1pt solid black;"
controls>
<source src="videos\1062_DFA_HAP_XX.mp4" type="video/mp4">
</video>
</div>
<div class="col s3 center">
<video style="width: 100%; border: 1pt solid black;"
controls>
<source src="videos\1082_WSI_SAD_XX.mp4" type="video/mp4">
</video>
</div>
</div>
<div class="row">
<h3 style="font-variant: small-caps;">
Generalization
</h3>
<p align="justify" style="font-size:18px">
Results of video generation given inputs from sources not used during the training.
</p>
<div class="col s4 center">
<video style="width: 100%; border: 1pt solid black;"
controls>
<source src="videos\lady.mp4" type="video/mp4">
</video>
<figcaption style="font-size: 1.4em;">
English female audio
</figcaption>
</div>
<div class="col s4 center">
<video style="width: 100%; border: 1pt solid black;"
controls>
<source src="videos\korean.mp4" type="video/mp4">
</video>
<figcaption style="font-size: 1.4em;">
Korean female audio
</figcaption>
</div>
<div class="col s4 center">
<video style="width: 100%; border: 1pt solid black;"
controls>
<source src="videos\dalle_sound.mp4" type="video/mp4">
</video>
<figcaption style="font-size: 1.4em;">
DALL-E 2 image + English male audio
</figcaption>
</div>
</div>
</div>
</div>
</div>
<hr>
<div class="row">
<h3>Abstract</h3>
<div class="row">
<div class="col s10 offset-s1">
<p align="justify" style="font-size:18px">
Talking face generation has historically struggled
to produce head movements and natural facial expressions
without guidance from additional reference videos. Recent
developments in diffusion-based generative models allow
for more realistic and stable data synthesis and their
performance on image and video generation has surpassed
that of other generative models. In this work, we present
an autoregressive diffusion model that requires only one
identity image and audio sequence to generate a video of
a realistic talking human head. Our solution is capable
of hallucinating head movements, facial expressions,
such as blinks, and preserving a given background.
We evaluate our model on two different datasets,
achieving state-of-the-art results on both of them.
</p>
</div>
</div>
<div class="row">
<h3>Bibtex</h3>
<pre>
@inproceedings{stypulkowski2022diffused,
title = {{Diffused Heads: Diffusion Models Beat GANs on Talking-Face Generation}},
author = {Stypu{\l}kowski, Micha{\l} and Vougioukas, Konstantinos and He, Sen and Zi\k{e}ba, Maciej and Petridis, Stavros and Pantic, Maja},
booktitle = {https://arxiv.org/abs/2301.03396},
year = {2023}
} </pre>
</div>
<div class="row">
<h3>Acknowledgements</h3>
<p>
We thank Kacper Kania for the <a href="https://github.com/conerf/conerf.github.io">website's template</a>.
</p>
</div>
</div>
<script src="https://code.jquery.com/jquery-3.3.1.slim.min.js"
integrity="sha384-q8i/X+965DzO0rT7abK41JStQIAqVgRVzpbzo5smXKp4YfRvH+8abtTE1Pi6jizo"
crossorigin="anonymous"></script>
<script
src="https://cdnjs.cloudflare.com/ajax/libs/popper.js/1.14.7/umd/popper.min.js"
integrity="sha384-UO2eT0CpHqdSJQ6hJty5KVphtPhzWj9WO1clHTMGa3JDZwrnQq4sF86dIHNDz0W1"
crossorigin="anonymous"></script>
<!-- JavaScript Bundle with Popper -->
<script
src="https://cdn.jsdelivr.net/npm/bootstrap@5.0.0-beta2/dist/js/bootstrap.bundle.min.js"
integrity="sha384-b5kHyXgcpbZJO/tY9Ul7kGkf1S0CWuKcCD38l8YkeH8z8QjE0GmW1gYU5S9FOnJ0"
crossorigin="anonymous"></script>
<script>
MathJax = {
tex: {
inlineMath: [
['$', '$'],
['\\(', '\\)']
]
},
svg: {
fontCache: 'global'
}
};
</script>
<script type="text/javascript" id="MathJax-script" async
src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-svg.js">
</script>
<script src="https://cdn.jsdelivr.net/npm/prismjs/prism.min.js">
</script>
<!-- Compiled and minified JavaScript -->
<script
src="https://cdnjs.cloudflare.com/ajax/libs/materialize/1.0.0/js/materialize.min.js">
</script>
<script src="https://kit.fontawesome.com/2f4b26bfb0.js"
crossorigin="anonymous"></script>
<script type="text/javascript">
document.addEventListener('DOMContentLoaded', function () {
var elems = document.querySelectorAll('.tooltipped');
var instances = M.Tooltip.init(elems, options);
});
// Or with jQuery
$(document).ready(function () {
$('.tooltipped').tooltip();
});
</script>
</body>
</html>