Skip to content
This repository was archived by the owner on Jan 10, 2023. It is now read-only.

Commit dd79f95

Browse files
shanyangmieshanyangmie
committed
text update
1 parent b7b10ca commit dd79f95

File tree

5 files changed

+462
-5
lines changed

5 files changed

+462
-5
lines changed

‎choreo_gen.html‎

Lines changed: 372 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,372 @@
1+
<!DOCTYPE html>
2+
<html lang="en"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
3+
<meta name="description" content="Project Page for Infinite Nature: Perpetual View Generation">
4+
<meta name="author" content="Andrew">
5+
6+
<meta property="og:image" content="https://infinite-nature.github.io/teaser.png">
7+
<meta property="og:url" content="https://infinite-nature.github.io/">
8+
<meta property="og:type" content="website">
9+
<meta property="og:title" content="Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image">
10+
<meta property='og:video' content='https://www.youtube.com/embed/oXUf6anNAtc'/>
11+
<meta property="og:description" content="Learning to generate arbitrarily long videos of aerial trajectories through nature scenes.">
12+
<meta name="twitter:card" content="summary_large_image">
13+
<meta name="twitter:creator" content="@ndrewliu">
14+
<meta name="twitter:title" content="Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image">
15+
<meta name="twitter:description" content="Learning to generate arbitrarily long videos of aerial trajectories through nature scenes.">
16+
<meta name=”twitter:url” content=”https://infinite-nature.github.io” />
17+
<meta name="twitter:image" content="https://infinite-nature.github.io/teaser.png">
18+
19+
<title>Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image</title>
20+
<link rel="canonical" href="https://infinite-nature.github.io">
21+
22+
<!-- Bootstrap core CSS -->
23+
<link href="./index_files/bootstrap.min.css" rel="stylesheet">
24+
25+
<head>
26+
<link href="https://fonts.googleapis.com/css?family=Roboto:300" rel="stylesheet">
27+
<link rel="stylesheet" type="text/css" href="main.css">
28+
<script type="text/javascript" src="main.js"></script>
29+
<!-- Global site tag (gtag.js) - Google Analytics -->
30+
<script async src="https://www.googletagmanager.com/gtag/js?id=G-JM2CPK6QLP"></script>
31+
<script>
32+
window.dataLayer = window.dataLayer || [];
33+
function gtag(){dataLayer.push(arguments);}
34+
gtag('js', new Date());
35+
36+
gtag('config', 'G-JM2CPK6QLP');
37+
</script>
38+
<script type="text/javascript">
39+
// Toggle Display of BibTeX
40+
function toggleBibtex(articleid) {
41+
var bib = document.getElementById('bib_'+articleid);
42+
if (bib) {
43+
if(bib.className.indexOf('bibtex') != -1) {
44+
bib.className.indexOf('noshow') == -1?bib.className = 'bibtex noshow':bib.className = 'bibtex';
45+
}
46+
} else {
47+
return;
48+
}
49+
}
50+
</script>
51+
52+
</head>
53+
54+
<style>
55+
/*************************************
56+
The box that contain BibTeX code
57+
*************************************/
58+
div.noshow { display: none; }
59+
div.bibtex {
60+
margin-right: 0%;
61+
margin-top: 1.2em;
62+
margin-bottom: 1em;
63+
border: 1px solid silver;
64+
padding: 0em 1em;
65+
background: #ffffee;
66+
}
67+
div.bibtex pre { font-size: 90%; overflow: auto; width: 100%; padding: 0em 0em;}
68+
69+
70+
.main {
71+
margin: 0 auto;
72+
padding: 15px 14px 50px;
73+
font-family: sans-serif;
74+
background: #ffffff;
75+
color: #222;
76+
max-width: 800px;
77+
}
78+
p {
79+
line-height: 150%;
80+
}
81+
h1, h2 {
82+
font-weight: 200;
83+
}
84+
h2 a {
85+
color: #000;
86+
}
87+
88+
h3 {
89+
margin-top: 40px;
90+
}
91+
a {
92+
text-decoration: none;
93+
color: #22b;
94+
}
95+
a:hover {
96+
text-decoration: underline;
97+
color: #22b;
98+
}
99+
a:hover img {
100+
opacity: .5;
101+
}
102+
.paper img { background: #888; }
103+
.video img {
104+
border-radius: 6px;
105+
background: #000;
106+
}
107+
h1 span {
108+
font-size: 90%;
109+
line-height: 1.5em;
110+
}
111+
h1 span::before { content: "("; }
112+
h1 span::after { content: ")"; }
113+
h1 span::before, h1 span::after {
114+
vertical-align: .06em;
115+
}
116+
h2 span {
117+
line-height: 2em;
118+
}
119+
em {
120+
color: #333;
121+
font-weight: bold;
122+
font-style: normal;
123+
}
124+
.teaser {
125+
max-width: 45em;
126+
margin: 40px 0 60px 0;
127+
white-space: nowrap;
128+
position: relative;
129+
}
130+
.teaser .imgs .full {
131+
width: 100%;
132+
}
133+
.teaser .imgs .center {
134+
position: absolute;
135+
width: 32%;
136+
top: -3%;
137+
left: 34%;
138+
}
139+
.teaser .labels span {
140+
display: inline-block;
141+
width: 33%;
142+
text-align: center;
143+
padding-top: 10px;
144+
}
145+
.bibtex {
146+
white-space: pre;
147+
font-family: monospace;
148+
line-height: 150%;
149+
background: #f3f3f3;
150+
padding: 10px;
151+
display: inline-block;
152+
border-radius: 4px;
153+
}
154+
.crow > * {
155+
vertical-align: middle;
156+
margin-right: 14px;
157+
}
158+
a.img {
159+
display: inline-block;
160+
}
161+
.links {
162+
display: inline-block;
163+
line-height: 150%;
164+
padding: 8px 0;
165+
}
166+
.links a {
167+
padding: 0 3px;
168+
}
169+
.examples {
170+
padding: 10px 0;
171+
max-width: 50em;
172+
}
173+
.examples img {
174+
display: inline-block;
175+
width: 192px;
176+
height: 108px;
177+
178+
}
179+
.video-container {
180+
position: relative;
181+
width: 80%;
182+
padding-bottom: 45%;
183+
}
184+
.video {
185+
position: absolute;
186+
top: 0;
187+
left: 0;
188+
width: 100%;
189+
height: 100%;
190+
border: 0;
191+
}
192+
</style>
193+
194+
<body onload="mainDivResize('choreo_gen')" onresize="mainDivResize()">
195+
196+
<div class="topnav" id="myTopnav">
197+
<a id="100" href="index.html" class="title">AIST++ Dataset</a>
198+
<!-- <div id="challenge" class="menu-dropdown">
199+
<a href="challenge_overview.html" class="droplink">Challenge</a>
200+
<div class="menu-dropdown-content" style="right:0px;">
201+
<a href="challenge_overview.html" class="subitem">Overview</a>
202+
<a href="challenge2019_downloads.html" class="subitem">Downloads</a>
203+
<a href="evaluation.html" class="subitem">Evaluation</a>
204+
<a href="challenge2019_guidelines.html" class="subitem">Participation guidelines</a>
205+
<a href="challenge2019.html" class="subitem">Past challenge: 2019</a>
206+
<a href="challenge.html" class="subitem">Past challenge: 2018</a>
207+
</div>
208+
</div> -->
209+
<!-- <a id="news" href="news.html" class="menuitem">News</a>
210+
<a id="extras" href="extras.html" class="menuitem">Extras</a>
211+
<a id="extended" href="extended.html" class="menuitem">Extended</a> -->
212+
<a id="team" href="team.html" class="menuitem">Team</a>
213+
<a id="explore" href="visualizer/index.html" class="menuitem">Explore</a>
214+
<a id="download" href="download.html" class="menuitem">Download</a>
215+
<a id="factsfigures" href="factsfigures.html" class="menuitem">Description</a>
216+
<a id="choreo_gen" href="choreo_gen.html" class="menuitem">Dance Generation</a>
217+
<a id="0" href="javascript:void(0);" style="font-size:15px;" class="icon" onclick="navbarResize()">&#9776;</a>
218+
219+
</div>
220+
<div class="main" id ="main">
221+
<div id="choreo_gen_banner">
222+
</div>
223+
</div>
224+
225+
<div class="main">
226+
<h1 align="center">Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image</h1><br>
227+
228+
<table align=center width=100%>
229+
<tr>
230+
<td align=center width=100px>
231+
<span style="font-size:16px"><a href="https://andrewhliu.github.io">Andrew Liu</a><sup>*</sup></span>
232+
</td>
233+
<td align=center width=100px>
234+
<span style="font-size:16px"><a href="https://research.google/people/RichardTucker/">Richard Tucker</a><sup>*</sup></span>
235+
</td>
236+
<td align=center width=100px>
237+
<span style="font-size:16px"><a href="https://varunjampani.github.io/">Varun Jampani</a></span>
238+
</td>
239+
</tr>
240+
<tr>
241+
<td align=center width=100px>
242+
<span style="font-size:16px"><a href="http://www.ameeshmakadia.com/index.html">Ameesh Makadia</a></span>
243+
</td>
244+
<td align=center width=100px>
245+
<span style="font-size:16px"><a href="http://www.cs.cornell.edu/~snavely/">Noah Snavely</a></span>
246+
</td>
247+
<td align=center width=100px>
248+
<span style="font-size:16px"><a href="https://people.eecs.berkeley.edu/~kanazawa/">Angjoo Kanazawa</a></span>
249+
</td>
250+
</tr>
251+
</table>
252+
<h5 align=center style="font-size:16px;font-weight:normal" >Google Research</h5>
253+
<hr>
254+
<br>
255+
<video autoplay loop muted playsinline width="100%">
256+
<source src="teaser_loop.mp4" type="video/mp4">
257+
</video>
258+
<table width=100%>
259+
<tr>
260+
<td align=center width=42.5%>
261+
<span style="font-size:16px">Input</span>
262+
</td>
263+
<td align=center width=15%>
264+
</td>
265+
<td align=center width=42.5%>
266+
<span style="font-size:16px">Generated Video</span>
267+
</td>
268+
</tr>
269+
</table>
270+
<!-- <table align=center width=100%>
271+
<tr>
272+
<td width=75%>
273+
</td>
274+
<td width=25%>
275+
<video autoplay loop muted playsinline width="100%">
276+
<source src="animation0.mp4" type="video/mp4">
277+
</video>
278+
</td>
279+
</table> -->
280+
281+
<h3>Abstract</h3>
282+
<p align="justify">We introduce the problem of <i>perpetual view generation</i>&mdash;long-range generation of novel views
283+
corresponding to an arbitrarily long camera trajectory given a single image. This is a
284+
challenging problem that goes far beyond the capabilities of current view synthesis methods,
285+
which work for a limited range of viewpoints and quickly degenerate when presented with a
286+
large camera motion. Methods designed for video generation also have limited ability to produce
287+
long video sequences and are often agnostic to scene geometry. We take a hybrid approach that
288+
integrates both geometry and image synthesis in an iterative <i>render</i>, <i>refine</i>, and <i>repeat</i> framework,
289+
allowing for long-range generation that cover large distances after hundreds of frames. Our approach
290+
can be trained from a set of monocular video sequences without any manual annotation. We propose a
291+
dataset of aerial footage of natural coastal scenes, and compare our method with recent view synthesis
292+
and conditional video generation baselines, showing that it can generate plausible scenes for much
293+
longer time horizons over large camera trajectories compared to existing methods.
294+
</p>
295+
296+
<h3>Paper</h3>
297+
<div class=crow>
298+
<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>
299+
<div class=links>
300+
<em>Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image</em><br>
301+
Andrew Liu*, Richard Tucker*, Varun Jampani,<br>
302+
Ameesh Makadia, Noah Snavely, Angjoo Kanazawa<br><br>
303+
arXiv<br>
304+
[<a href="http://arxiv.org/abs/2012.09855">Arxiv</a>]
305+
</div>
306+
</div>
307+
308+
<h3>Video</h3>
309+
<div class="video-container">
310+
<iframe class="video" src="https://www.youtube.com/embed/oXUf6anNAtc" allowfullscreen></iframe>
311+
</div>
312+
<div class=crow>
313+
<div class=links>
314+
[<a href="https://youtu.be/oXUf6anNAtc">YouTube</a>]
315+
</div>
316+
</div>
317+
318+
<h3>Code</h3>
319+
<p>Coming Soon!</p>
320+
321+
<h3>Perpetual View Generation</h3>
322+
323+
<div align="center">
324+
<video autoplay loop muted playsinline controls width="32%">
325+
<source src="animation0.mp4" type="video/mp4">
326+
</video>
327+
<video autoplay loop muted playsinline controls width="32%">
328+
<source src="animation1.mp4" type="video/mp4">
329+
</video>
330+
<video autoplay loop muted playsinline controls width="32%">
331+
<source src="animation4.mp4" type="video/mp4">
332+
</video>
333+
<br>
334+
<video autoplay loop muted playsinline controls width="32%">
335+
<source src="animation2.mp4" type="video/mp4">
336+
</video>
337+
<video autoplay loop muted playsinline controls width="32%">
338+
<source src="animation3.mp4" type="video/mp4">
339+
</video>
340+
</div>
341+
342+
343+
<h3>Aerial Coastline Imagery Dataset (ACID)</h3>
344+
<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>
345+
<br>
346+
<div align=center>
347+
<iframe src="https://www.youtube.com/embed/3gYsoFlJzMo?mute=1" width=32% frameborder="0"> </iframe>
348+
<iframe src="https://www.youtube.com/embed/Ow1kqyTRsvY?mute=1" width=32% frameborder="0"> </iframe>
349+
<iframe src="https://www.youtube.com/embed/B5pl4yRKmug?mute=1" width=32% frameborder="0"> </iframe>
350+
<br>
351+
<iframe src="https://www.youtube.com/embed/zcW7DealMxU?mute=1" width=32% frameborder="0"> </iframe>
352+
<iframe src="https://www.youtube.com/embed/ySmTXPeF0Sg?mute=1" width=32% frameborder="0"> </iframe>
353+
<iframe src="https://www.youtube.com/embed/xHstpLRmhWg?mute=1" width=32% frameborder="0"> </iframe>
354+
</div>
355+
<br>
356+
[<a href="http://storage.googleapis.com/gresearch/aerial-coastline-imagery-dataset/acid_v1_release.tar.gz">Download Dataset</a>]
357+
358+
<h3>BibTeX</h3>
359+
<div class=bibtex>@InProceedings{infinite_nature_2020,
360+
author = {Liu, Andrew and Tucker, Richard and Jampani, Varun and
361+
Makadia, Ameesh and Snavely, Noah and Kanazawa, Angjoo},
362+
title = {Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image},
363+
booktitle = {arXiv},
364+
month = {December},
365+
year = {2020}
366+
}
367+
</div>
368+
369+
</div>
370+
371+
</body>
372+
</html>

‎factsfigures.html‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -100,10 +100,10 @@ <h2>Publications</h2>
100100
<img src="images/paper_teaser.gif">
101101
</td>
102102
<td width=75% style="border: 0;">
103-
<i>Ruilong Li*, Shan Yang*, David Ross, Angjoo Kanazawa.</i><br>
103+
<i>Ruilong Li*, Shan Yang*, David A. Ross, Angjoo Kanazawa.</i><br>
104104
Learn to Dance with AIST++: Music Conditioned 3D Dance Generation.<br>
105-
In submission, 2020.<br>
106-
<a href="https://arxiv.org/abs/TODO" target="_blank">[PDF]</a> <a href="javascript:toggleBibtex('DanceGen')">[BibTeX]</a>
105+
arXiv, 2021.<br>
106+
<a href="https://arxiv.org/abs/TODO" target="_blank">[PDF]</a> <a href="javascript:toggleBibtex('DanceGen')">[BibTeX]</a> <a href="https://google.github.io/aichoreographer">[Web]</a>
107107
</td>
108108
</tr>
109109
</table>
@@ -135,7 +135,7 @@ <h2>Dataset organization</h2>
135135
<p>The dataset is split into training/validation/testing sets in different ways serving for different purposes.</p>
136136
<ul>
137137
<li>For tasks such as human pose estimation and human motion prediction, we recommend using data splits described in Table 1. Here we split the trainval and testing sets based on different <em>subjects</em>, which also makes sure the human motions in the trainval and testing sets have no overlap. Note training and validation sets share the same group of subjects.</li>
138-
<li>For tasks dealing with motion and music such as music conditioned motion generation, we recommend using data splits described in Table 2. In the AIST database, the same music and the same choreography are shared by multiple human motion sequences. So we carefully split the dataset to make sure that the music and choreography in the training set does not overlap with that in the testing/validation set. Note validation and testing sets share the same group of music but different choreographies.</li>
138+
<li>For tasks dealing with motion and music such as music conditioned motion generation, we recommend using data splits described in Table 2. In the AIST database, same music and same choreography are shared by multiple human motion sequences. So we carefully split the dataset to make sure that the music and choreography in the training set does not overlap with that in the testing/validation set. Note validation and testing sets share the same group of music but different choreographies.</li>
139139
</ul>
140140
<p align='center'>Table 1: Data Splits based on Subjects.</p>
141141
<table>

0 commit comments

Comments
 (0)