-
Notifications
You must be signed in to change notification settings - Fork 14
/
index.html
477 lines (440 loc) · 22.1 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
<!doctype html>
<html lang="en">
<!-- === Header Starts === -->
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<meta name="description" content="HyperHuman Project Page">
<title>HyperHuman: Hyper-Realistic Human Generation with Latent Structural Diffusion</title>
<link href="./static/css/bootstrap.min.css" rel="stylesheet">
<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">
<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 href="./static/css/font.css" rel="stylesheet" type="text/css">
<link href="./static/css/style.css" rel="stylesheet" type="text/css">
</head>
<!-- === Header Ends === -->
<body>
<!-- === Home Section Starts === -->
<div class="section">
<!-- === Title Starts === -->
<div class="header">
<!-- <div class="logo">
<a href="https://mmlab.ie.cuhk.edu.hk/" target="_blank"><img src="./common/cuhk.jfif"></a>
</div> -->
<div class="title", style="font-size: 24pt; padding-top: 10pt;"> <!-- Set padding as 10 if title is with two lines. -->
HyperHuman: Hyper-Realistic Human Generation with Latent Structural Diffusion
<!-- <br>
<font color="grey" size="4">arXiv Preprint.</font> -->
</div>
</div>
<!-- === Title Ends === -->
<div class="author">
<a href="https://alvinliu0.github.io/" target="_blank">Xian Liu</a><sup>1,2</sup>
<a href="https://alanspike.github.io/" target="_blank">Jian Ren</a><sup>1</sup>
<a href="https://aliaksandrsiarohin.github.io/aliaksandr-siarohin-website/" target="_blank">Aliaksandr Siarohin</a><sup>1</sup>
<a href="https://universome.github.io/" target="_blank">Ivan Skorokhodov</a><sup>1</sup>
<a href="https://scholar.google.com/citations?user=XUj8koUAAAAJ&hl=en" target="_blank">Yanyu Li</a><sup>1</sup>
<br>
<a href="http://dahua.site/" target="_blank">Dahua Lin</a><sup>2</sup>
<a href="https://xh-liu.github.io/" target="_blank">Xihui Liu</a><sup>3</sup>
<a href="https://liuziwei7.github.io/" target="_blank">Ziwei Liu</a><sup>4</sup>
<a href="http://www.stulyakov.com/" target="_blank">Sergey Tulyakov</a><sup>1</sup>
</div>
<div class="institution">
<sup>1</sup>Snap Inc.
<sup>2</sup>CUHK
<sup>3</sup>HKU
<sup>4</sup>NTU
</div>
<div class="column has-text-centered">
<div class="publication-links">
<!-- PDF Link. -->
<span class="link-block">
<a href="./content/hyperhuman.pdf" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Paper</span>
</a>
</span>
<span class="link-block">
<a href="https://arxiv.org/abs/2310.08579" 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>
<!-- Video Link. -->
<span class="link-block">
<a href="https://www.youtube.com/watch?v=eRPZW1pwxog"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-youtube"></i>
</span>
<span>Short Demo (3min)</span>
</a>
</span>
<span class="link-block">
<a href="https://www.youtube.com/watch?v=CxGfbwZOcyU"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-youtube"></i>
</span>
<span>Long Demo (10min)</span>
</a>
</span>
<!-- Code Link. -->
<span class="link-block">
<a href="https://github.com/snap-research/HyperHuman"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Github</span>
</a>
</span>
<!-- Dataset Link.
<span class="link-block">
<a href="./#BibTeX" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fa fa-quote-left"></i>
</span>
<span>Cite</span>
</a> -->
</div>
</div>
<div class="container" width=1000px>
<div id="results-carousel" class="carousel results-carousel">
<div class="card-image" style="text-align: center;">
<img src="./content/examples/baby girl.png" width="80%">
<div class="card-content has-text-centered">
A baby girl with beautiful blue eyes standing next to a brown teddy bear.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/little girl.png" width="80%">
<div class="card-content has-text-centered">
A little girl with wavy hair and smile holding a teddy bear.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/A man and woman.png" width="80%">
<div class="card-content has-text-centered">
A man and woman seated at a table in a restaurant.
</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/A cow laying.png" width="80%">
<div class="card-content has-text-centered">
A cow laying on the grass behind a man holding a cup of coffee.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/A young kid.png" width="80%">
<div class="card-content has-text-centered">
A young kid stands before a birthday cake decorated with captain America.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/A man who is sitting.png" width="80%">
<div class="card-content has-text-centered">
A man who is sitting in a bus looking away from the window.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/a man in a red.png" width="80%">
<div class="card-content has-text-centered">
A man in a red shirt is holding a skate board up over his head.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/two mean who are.png" width="80%">
<div class="card-content has-text-centered">
Two men who are sitting next to each other with a large pizza in front of them.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/two children.png" width="80%">
<div class="card-content has-text-centered">
Two children carry an enormous stuffed teddy bear.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/the upper half.png" width="80%">
<div class="card-content has-text-centered">
The upper half of a man posing for a photograph wearing a suit with a blue tie and matching pocket corner.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/an older man.png" width="80%">
<div class="card-content has-text-centered">
An older man is wearing a funny hat in his dining room.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/young man.png" width="80%">
<div class="card-content has-text-centered">
Young man on top of a snowboard wearing maroon jacket.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/man seating.png" width="80%">
<div class="card-content has-text-centered">
Man sitting on brick covered ground, appearing dirty and tired.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/a man wearing.png" width="80%">
<div class="card-content has-text-centered">
A man wearing a purple neck tie and glasses while sitting in a car.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/a man standing.png" width="80%">
<div class="card-content has-text-centered">
A man standing on grassy area next to trees.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/a girl with blue.png" width="80%">
<div class="card-content has-text-centered">
A girl with blue hair is taking a self portrait.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/a man wearing a helmet.png" width="80%">
<div class="card-content has-text-centered">
A man wearing a helmet is sitting on his blue motorcycle.</div>
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/examples/a person dressed up.png" width="80%">
<div class="card-content has-text-centered">
A person dressed up taking a picture at a street with his fist up.</div>
</div>
<!-- <div class="item item-clevr" style="text-align: center;">
<img src="./content/our_more3.png" width="90%">
</div> -->
</div>
</div>
<!-- <div class="link">
<a href="https://arxiv.org/pdf/2212.02350.pdf" target="_blank">[Paper]</a>
<a href="https://github.com/alvinliu0/ANGIE" target="_blank">[Code]</a>
<a href="https://github.com/alvinliu0/ANGIE" target="_blank">[Dataset]</a>
</div> -->
</div>
<!-- === Home Section Ends === -->
<!-- === Result Section Starts === -->
<!-- === Result Section Ends === -->
<div class="section">
<div class="title">Short Demo Video (3min)</div>
<div class="body">
We present a short demo video, mostly with visualization results and a very quick overview of our framework.
<div class="publication-video">
<iframe src="https://www.youtube.com/embed/eRPZW1pwxog" title="YouTube video player"
frameborder="0"
allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture"
allowfullscreen></iframe>
</div>
</div>
</div>
<div class="section">
<div class="title">Long Demo Video (10min)</div>
<div class="body">
We present a long demo video with detailed elaborations on the motivations and framework designs.
<div class="publication-video">
<iframe src="https://www.youtube.com/embed/CxGfbwZOcyU" title="YouTube video player"
frameborder="0"
allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture"
allowfullscreen></iframe>
</div>
</div>
</div>
<!-- === Overview Section Starts === -->
<div class="section">
<div class="title">Abstract</div>
<div class="body">
Despite significant advances in large-scale text-to-image models, achieving hyper-realistic human image generation remains a desirable yet unsolved task.
Existing models like Stable Diffusion and DALL·E 2 tend to generate human images with incoherent parts or unnatural poses.
To tackle these challenges, our key insight is that human image is inherently structural over multiple granularities, from the coarse-level body skeleton to the fine-grained spatial geometry.
Therefore, capturing such correlations between the explicit appearance and latent structure in one model is essential to generate coherent and natural human images.
To this end, we propose a unified framework, <b>HyperHuman</b>, that generates in-the-wild human images of high realism and diverse layouts.
Specifically, <b>1)</b> we first build a large-scale human-centric dataset, named <i>HumanVerse</i>, which consists of 340M images with comprehensive annotations like human pose, depth, and surface-normal.
<b>2)</b> Next, we propose a <i>Latent Structural Diffusion Model</i> that simultaneously denoises the depth and surface-normal along with the synthesized RGB image. Our model enforces the joint learning of image appearance, spatial relationship, and geometry in a unified network, where each branch in the model complements to each other with both structural awareness and textural richness.
<b>3)</b> Finally, to further boost the visual quality, we propose a <i>Structure-Guided Refiner</i> to compose the predicted conditions for more detailed generation of higher resolution.
Extensive experiments demonstrate that our framework yields the state-of-the-art performance, generating hyper-realistic human images under diverse scenarios.
</div>
<div class="container" width=1000px>
<div id="teaser-carousel" class="carousel2 results-carousel">
<div class="card-image" style="text-align: center;">
<img src="./content/teaser/t1.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/teaser/t2.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/teaser/t3.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/teaser/t4.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/teaser/t5.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/teaser/t6.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/teaser/t7.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/teaser/t8.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/teaser/t9.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/teaser/t10.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/teaser/t11.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/teaser/t12.jpg" width="90%">
</div>
</div>
</div>
<div class="teaser">
<img src="./content/teaser_half.png">
</div>
<i>Top:</i> The proposed <b>HyperHuman</b> simultaneously generates the coarse RGB, depth, normal, and high-resolution images conditioned on text and skeleton.
Both photo-realistic images and stylistic renderings can be created. <i>Bottom:</i> We compare with recent T2I models, showing better realism, quality, diversity, and controllability.
Note that in each 2x2 grid (<b>left</b>), the upper-left is <i>input</i> skeleton, while the others are jointly denoised normal, depth, and coarse RGB of 512x512.
With full model, we synthesize images up to 1024x1024 (<b>right</b>).
</div>
<!-- === Overview Section Ends === -->
<div class="section">
<div class="title">Framework Overview</div>
<div class="body">
<!-- Adjust the number of rows and columns (EVERY project differs). -->
<table width="100%" style="margin: 20pt 0; text-align: center;">
<tr>
<td><img src="./content/framework.png" width="90%"></td>
</tr>
</table>
<b>Overview of HyperHuman Framework.</b> In <i>Latent Structural Diffusion Model</i> (<span style="color: rgb(230, 128, 230); white-space: nowrap">purple</span>), the image <b>x</b>, depth <b>d</b>, and surface-normal <b>n</b> are jointly denoised conditioning on caption <b>c</b> and pose skeleton <b>p</b>. In <i>Structure-Guided Refiner</i> (<span style="color: rgb(46, 118, 181); white-space: nowrap">blue</span>), we compose the predicted conditions for higher-resolution generation. Note that the grey images refer to randomly dropout conditions for more robust training.
</div>
</div>
<!-- === Result Section Starts === -->
<div class="section">
<div class="title">Quantitative Results</div>
<div class="body">
<!-- Adjust the number of rows and columns (EVERY project differs). -->
<table width="100%" style="margin: 20pt 0; text-align: center;">
<tr>
<td><img src="./content/quantitative1.png" width="90%"></td>
</tr>
</table>
<b>Zero-Shot Evaluation on MS-COCO 2014 Validation Human.</b> We compare our model with recent SOTA general T2I models (Stable Diffusion v1.5, v2.0, v2.1; SDXL; DeepFloyd-IF) and controllable methods (ControlNet; T2I-Adapter; HumanSD). Note that SDXL generates artistic style in 512x512, and IF only creates fixed-size images, we first generate 1024x1024 results, then resize back to 512x512 for these two methods. We bold the <b>best</b> and underline the <u>second</u> results for clarity. Our improvements over the second method are shown in <span style="color: rgb(255, 0, 0); white-space: nowrap">red</span>.
<br>
<br>
<table width="100%" style="margin: 20pt 0; text-align: center;">
<tr>
<td><img src="./content/quantitative2.png" width="90%"></td>
</tr>
</table>
<b>Evaluation Curves on MS-COCO 2014 Validation Human Subset.</b> We show FID-CLIP (<i>left</i>) and FID<sub>CLIP</sub>-CLIP (<i>right</i>) curves with CFG scale ranging from 4.0 to 20.0 for all methods.
<br>
<br>
<table width="100%" style="margin: 20pt 0; text-align: center;">
<tr>
<td><img src="./content/quantitative3.png" width="90%"></td>
</tr>
</table>
<b>User Preference Comparisons.</b> We report the ratio of users prefer our model to baselines.
</div>
</div>
<!-- === Result Section Ends === -->
<!-- === Result Section Starts === -->
<div class="section">
<div class="title">More Comparisons (1024x1024)</div>
<div class="body">
<!-- Adjust the number of rows and columns (EVERY project differs). -->
<!-- <table width="100%" style="margin: 20pt 0; text-align: center;"> -->
<div class="container" width=1000px>
<div id="results-carousel2" class="carousel2 results-carousel">
<div class="card-image" style="text-align: center;">
<img src="./content/comparison/com1.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/comparison/com2.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/comparison/com3.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/comparison/com4.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/comparison/com5.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/comparison/com6.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/comparison/com7.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/comparison/com8.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/comparison/com9.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/comparison/com10.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/comparison/com11.jpg" width="90%">
</div>
<div class="card-image" style="text-align: center;">
<img src="./content/comparison/com12.jpg" width="90%">
</div>
</div>
</div>
<!-- </table> -->
</div>
</div>
<!-- === Result Section Ends === -->
<!-- === Result Section Starts === -->
<!-- <div class="section">
<div class="title">More Qualitative Results (1024x1024)</div>
<div class="body"> -->
<!-- Adjust the number of rows and columns (EVERY project differs). -->
<!-- <table width="100%" style="margin: 20pt 0; text-align: center;">
<div class="container" width=1000px>
<div id="results-carousel" class="carousel results-carousel">
<div class="item item-clevr" style="text-align: center;">
<img src="./content/our_more1.png" width="90%">
</div>
<div class="item item-clevr" style="text-align: center;">
<img src="./content/our_more2.png" width="90%">
</div>
<div class="item item-clevr" style="text-align: center;">
<img src="./content/our_more3.png" width="90%">
</div>
</div>
</div> -->
<!-- <tr>
<td><img src="./content/our_more1.png" width="90%"></td>
</tr> -->
<!-- </table>
</div>
</div> -->
<!-- === Result Section Ends === -->
<!-- === Reference Section Starts === -->
<div class="section">
<div class="bibtex">BibTeX</div>
<pre>
@article{liu2023hyperhuman,
title={HyperHuman: Hyper-Realistic Human Generation with Latent Structural Diffusion},
author={Liu, Xian and Ren, Jian and Siarohin, Aliaksandr and Skorokhodov, Ivan and Li, Yanyu and Lin, Dahua and Liu, Xihui and Liu, Ziwei and Tulyakov, Sergey},
journal={arXiv preprint arXiv:2310.08579},
year={2023}
}
</pre>
</body>
</html>