This repository has been archived by the owner on Jan 10, 2023. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 4
/
choreo_gen.html
372 lines (345 loc) · 12.1 KB
/
choreo_gen.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
<!DOCTYPE html>
<html lang="en"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<meta name="description" content="Project Page for Infinite Nature: Perpetual View Generation">
<meta name="author" content="Andrew">
<meta property="og:image" content="https://infinite-nature.github.io/teaser.png">
<meta property="og:url" content="https://infinite-nature.github.io/">
<meta property="og:type" content="website">
<meta property="og:title" content="Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image">
<meta property='og:video' content='https://www.youtube.com/embed/oXUf6anNAtc'/>
<meta property="og:description" content="Learning to generate arbitrarily long videos of aerial trajectories through nature scenes.">
<meta name="twitter:card" content="summary_large_image">
<meta name="twitter:creator" content="@ndrewliu">
<meta name="twitter:title" content="Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image">
<meta name="twitter:description" content="Learning to generate arbitrarily long videos of aerial trajectories through nature scenes.">
<meta name=”twitter:url” content=”https://infinite-nature.github.io” />
<meta name="twitter:image" content="https://infinite-nature.github.io/teaser.png">
<title>Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image</title>
<link rel="canonical" href="https://infinite-nature.github.io">
<!-- Bootstrap core CSS -->
<link href="./index_files/bootstrap.min.css" rel="stylesheet">
<head>
<link href="https://fonts.googleapis.com/css?family=Roboto:300" rel="stylesheet">
<link rel="stylesheet" type="text/css" href="main.css">
<script type="text/javascript" src="main.js"></script>
<!-- Global site tag (gtag.js) - Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-JM2CPK6QLP"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'G-JM2CPK6QLP');
</script>
<script type="text/javascript">
// Toggle Display of BibTeX
function toggleBibtex(articleid) {
var bib = document.getElementById('bib_'+articleid);
if (bib) {
if(bib.className.indexOf('bibtex') != -1) {
bib.className.indexOf('noshow') == -1?bib.className = 'bibtex noshow':bib.className = 'bibtex';
}
} else {
return;
}
}
</script>
</head>
<style>
/*************************************
The box that contain BibTeX code
*************************************/
div.noshow { display: none; }
div.bibtex {
margin-right: 0%;
margin-top: 1.2em;
margin-bottom: 1em;
border: 1px solid silver;
padding: 0em 1em;
background: #ffffee;
}
div.bibtex pre { font-size: 90%; overflow: auto; width: 100%; padding: 0em 0em;}
.main {
margin: 0 auto;
padding: 15px 14px 50px;
font-family: sans-serif;
background: #ffffff;
color: #222;
max-width: 800px;
}
p {
line-height: 150%;
}
h1, h2 {
font-weight: 200;
}
h2 a {
color: #000;
}
h3 {
margin-top: 40px;
}
a {
text-decoration: none;
color: #22b;
}
a:hover {
text-decoration: underline;
color: #22b;
}
a:hover img {
opacity: .5;
}
.paper img { background: #888; }
.video img {
border-radius: 6px;
background: #000;
}
h1 span {
font-size: 90%;
line-height: 1.5em;
}
h1 span::before { content: "("; }
h1 span::after { content: ")"; }
h1 span::before, h1 span::after {
vertical-align: .06em;
}
h2 span {
line-height: 2em;
}
em {
color: #333;
font-weight: bold;
font-style: normal;
}
.teaser {
max-width: 45em;
margin: 40px 0 60px 0;
white-space: nowrap;
position: relative;
}
.teaser .imgs .full {
width: 100%;
}
.teaser .imgs .center {
position: absolute;
width: 32%;
top: -3%;
left: 34%;
}
.teaser .labels span {
display: inline-block;
width: 33%;
text-align: center;
padding-top: 10px;
}
.bibtex {
white-space: pre;
font-family: monospace;
line-height: 150%;
background: #f3f3f3;
padding: 10px;
display: inline-block;
border-radius: 4px;
}
.crow > * {
vertical-align: middle;
margin-right: 14px;
}
a.img {
display: inline-block;
}
.links {
display: inline-block;
line-height: 150%;
padding: 8px 0;
}
.links a {
padding: 0 3px;
}
.examples {
padding: 10px 0;
max-width: 50em;
}
.examples img {
display: inline-block;
width: 192px;
height: 108px;
}
.video-container {
position: relative;
width: 80%;
padding-bottom: 45%;
}
.video {
position: absolute;
top: 0;
left: 0;
width: 100%;
height: 100%;
border: 0;
}
</style>
<body onload="mainDivResize('choreo_gen')" onresize="mainDivResize()">
<div class="topnav" id="myTopnav">
<a id="100" href="index.html" class="title">AIST++ Dataset</a>
<!-- <div id="challenge" class="menu-dropdown">
<a href="challenge_overview.html" class="droplink">Challenge</a>
<div class="menu-dropdown-content" style="right:0px;">
<a href="challenge_overview.html" class="subitem">Overview</a>
<a href="challenge2019_downloads.html" class="subitem">Downloads</a>
<a href="evaluation.html" class="subitem">Evaluation</a>
<a href="challenge2019_guidelines.html" class="subitem">Participation guidelines</a>
<a href="challenge2019.html" class="subitem">Past challenge: 2019</a>
<a href="challenge.html" class="subitem">Past challenge: 2018</a>
</div>
</div> -->
<!-- <a id="news" href="news.html" class="menuitem">News</a>
<a id="extras" href="extras.html" class="menuitem">Extras</a>
<a id="extended" href="extended.html" class="menuitem">Extended</a> -->
<a id="team" href="team.html" class="menuitem">Team</a>
<a id="explore" href="visualizer/index.html" class="menuitem">Explore</a>
<a id="download" href="download.html" class="menuitem">Download</a>
<a id="factsfigures" href="factsfigures.html" class="menuitem">Description</a>
<a id="choreo_gen" href="choreo_gen.html" class="menuitem">Dance Generation</a>
<a id="0" href="javascript:void(0);" style="font-size:15px;" class="icon" onclick="navbarResize()">☰</a>
</div>
<div class="main" id ="main">
<div id="choreo_gen_banner">
</div>
</div>
<div class="main">
<h1 align="center">Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image</h1><br>
<table align=center width=100%>
<tr>
<td align=center width=100px>
<span style="font-size:16px"><a href="https://andrewhliu.github.io">Andrew Liu</a><sup>*</sup></span>
</td>
<td align=center width=100px>
<span style="font-size:16px"><a href="https://research.google/people/RichardTucker/">Richard Tucker</a><sup>*</sup></span>
</td>
<td align=center width=100px>
<span style="font-size:16px"><a href="https://varunjampani.github.io/">Varun Jampani</a></span>
</td>
</tr>
<tr>
<td align=center width=100px>
<span style="font-size:16px"><a href="http://www.ameeshmakadia.com/index.html">Ameesh Makadia</a></span>
</td>
<td align=center width=100px>
<span style="font-size:16px"><a href="http://www.cs.cornell.edu/~snavely/">Noah Snavely</a></span>
</td>
<td align=center width=100px>
<span style="font-size:16px"><a href="https://people.eecs.berkeley.edu/~kanazawa/">Angjoo Kanazawa</a></span>
</td>
</tr>
</table>
<h5 align=center style="font-size:16px;font-weight:normal" >Google Research</h5>
<hr>
<br>
<video autoplay loop muted playsinline width="100%">
<source src="teaser_loop.mp4" type="video/mp4">
</video>
<table width=100%>
<tr>
<td align=center width=42.5%>
<span style="font-size:16px">Input</span>
</td>
<td align=center width=15%>
</td>
<td align=center width=42.5%>
<span style="font-size:16px">Generated Video</span>
</td>
</tr>
</table>
<!-- <table align=center width=100%>
<tr>
<td width=75%>
</td>
<td width=25%>
<video autoplay loop muted playsinline width="100%">
<source src="animation0.mp4" type="video/mp4">
</video>
</td>
</table> -->
<h3>Abstract</h3>
<p align="justify">We introduce the problem of <i>perpetual view generation</i>—long-range generation of novel views
corresponding to an arbitrarily long camera trajectory given a single image. This is a
challenging problem that goes far beyond the capabilities of current view synthesis methods,
which work for a limited range of viewpoints and quickly degenerate when presented with a
large camera motion. Methods designed for video generation also have limited ability to produce
long video sequences and are often agnostic to scene geometry. We take a hybrid approach that
integrates both geometry and image synthesis in an iterative <i>render</i>, <i>refine</i>, and <i>repeat</i> framework,
allowing for long-range generation that cover large distances after hundreds of frames. Our approach
can be trained from a set of monocular video sequences without any manual annotation. We propose a
dataset of aerial footage of natural coastal scenes, and compare our method with recent view synthesis
and conditional video generation baselines, showing that it can generate plausible scenes for much
longer time horizons over large camera trajectories compared to existing methods.
</p>
<h3>Paper</h3>
<div class=crow>
<a class=img href="http://arxiv.org/abs/2012.09855"><img src="inf_nat_first_page.png" height=150 style="border: 1px solid #555"></a>
<div class=links>
<em>Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image</em><br>
Andrew Liu*, Richard Tucker*, Varun Jampani,<br>
Ameesh Makadia, Noah Snavely, Angjoo Kanazawa<br><br>
arXiv<br>
[<a href="http://arxiv.org/abs/2012.09855">Arxiv</a>]
</div>
</div>
<h3>Video</h3>
<div class="video-container">
<iframe class="video" src="https://www.youtube.com/embed/oXUf6anNAtc" allowfullscreen></iframe>
</div>
<div class=crow>
<div class=links>
[<a href="https://youtu.be/oXUf6anNAtc">YouTube</a>]
</div>
</div>
<h3>Code</h3>
<p>Coming Soon!</p>
<h3>Perpetual View Generation</h3>
<div align="center">
<video autoplay loop muted playsinline controls width="32%">
<source src="animation0.mp4" type="video/mp4">
</video>
<video autoplay loop muted playsinline controls width="32%">
<source src="animation1.mp4" type="video/mp4">
</video>
<video autoplay loop muted playsinline controls width="32%">
<source src="animation4.mp4" type="video/mp4">
</video>
<br>
<video autoplay loop muted playsinline controls width="32%">
<source src="animation2.mp4" type="video/mp4">
</video>
<video autoplay loop muted playsinline controls width="32%">
<source src="animation3.mp4" type="video/mp4">
</video>
</div>
<h3>Aerial Coastline Imagery Dataset (ACID)</h3>
<p align=justify>In order to train our model, we identified thousands of aerial drone videos of different coastline and nature scenes on YouTube. We run structure-from-motion to get camera poses and release this data in the same format as <a href="https://google.github.io/realestate10k/">RealEstate10k</a>. Shown below are some randomly selected example videos that we identified.</p>
<br>
<div align=center>
<iframe src="https://www.youtube.com/embed/3gYsoFlJzMo?mute=1" width=32% frameborder="0"> </iframe>
<iframe src="https://www.youtube.com/embed/Ow1kqyTRsvY?mute=1" width=32% frameborder="0"> </iframe>
<iframe src="https://www.youtube.com/embed/B5pl4yRKmug?mute=1" width=32% frameborder="0"> </iframe>
<br>
<iframe src="https://www.youtube.com/embed/zcW7DealMxU?mute=1" width=32% frameborder="0"> </iframe>
<iframe src="https://www.youtube.com/embed/ySmTXPeF0Sg?mute=1" width=32% frameborder="0"> </iframe>
<iframe src="https://www.youtube.com/embed/xHstpLRmhWg?mute=1" width=32% frameborder="0"> </iframe>
</div>
<br>
[<a href="http://storage.googleapis.com/gresearch/aerial-coastline-imagery-dataset/acid_v1_release.tar.gz">Download Dataset</a>]
<h3>BibTeX</h3>
<div class=bibtex>@InProceedings{infinite_nature_2020,
author = {Liu, Andrew and Tucker, Richard and Jampani, Varun and
Makadia, Ameesh and Snavely, Noah and Kanazawa, Angjoo},
title = {Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image},
booktitle = {arXiv},
month = {December},
year = {2020}
}
</div>
</div>
</body>
</html>