-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
215 lines (196 loc) · 11.5 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
<html>
<head>
<!-- Google tag (gtag.js) -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-FYQNRK8LHK"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag() { dataLayer.push(arguments); }
gtag('js', new Date());
gtag('config', 'G-FYQNRK8LHK');
</script>
<meta charset="utf-8" />
<title>Incrementally Building Room-Scale Language-Embedded Gaussian Splats (LEGS) with a Mobile Robot</title>
<link href="https://fonts.googleapis.com" rel="preconnect" />
<link href="https://fonts.gstatic.com" rel="preconnect" crossorigin="anonymous" />
<script src="https://ajax.googleapis.com/ajax/libs/webfont/1.6.26/webfont.js" type="text/javascript"></script>
<script
type="text/javascript">WebFont.load({ google: { families: ["Lato:100,100italic,300,300italic,400,400italic,700,700italic,900,900italic", "Montserrat:100,100italic,200,200italic,300,300italic,400,400italic,500,500italic,600,600italic,700,700italic,800,800italic,900,900italic", "Ubuntu:300,300italic,400,400italic,500,500italic,700,700italic", "Open Sans:300,300italic,400,400italic,600,600italic,700,700italic,800,800italic", "Changa One:400,400italic", "Varela Round:400", "Bungee Shade:regular", "Roboto:300,regular,500", "Bungee Outline:regular"] } });</script>
<!--[if lt IE 9]><script src="https://cdnjs.cloudflare.com/ajax/libs/html5shiv/3.7.3/html5shiv.min.js" type="text/javascript"></script><![endif]-->
<script src="https://ajax.googleapis.com/ajax/libs/jquery/1/jquery.min.js"></script>
<script src="script.js" type="text/javascript"></script>
<link href="style.css" rel="stylesheet" type="text/css" />
<link href="data/favicon.png" rel="shortcut icon" type="image/x-icon" />
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/4.7.0/css/font-awesome.min.css">
</head>
<body>
<div class="section">
<div class="container">
<div class="title-row">
<h1 class="title">LEGS</h1>
<h1 class="subheader"> Incrementally Building Room-Scale Language-Embedded Gaussian Splats (LEGS) with a Mobile Robot </h1>
</div>
<div class="base-row author-row">
<div class="base-col author-col">
<a href="https://uynitsuj.github.io/" target="_blank" class="author-text">
Justin Yu
<span class="text-star">* 1</span>
<!-- <span class="text-star">*</span> -->
</a>
</div>
<div class="base-col author-col">
<a href="" target="_blank" class="author-text">
Kush Hari
<span class="text-star">* 1</span>
<!-- <span class="text-star">*</span> -->
</a>
</div>
<div class="base-col author-col">
<a href="" target="_blank" class="author-text">
Kishore Srinivas
<span class="text-star">* 1</span>
<!-- <span class="text-star">*</span> -->
</a>
</div>
<div class="base-col author-col">
<a href="" target="_blank" class="author-text">
Karim El-Refai
<span class="text-star">1</span>
<!-- <span class="text-star">*</span> -->
</a>
</div>
<div class="base-col author-col">
<a href="" target="_blank" class="author-text">
Adam Rashid
<span class="text-star">1</span>
</a>
</div>
</div>
<div class="base-row author-row">
<div class="base-col author-col">
<a href="" target="_blank" class="author-text">
Chung Min Kim
<span class="text-star">1</span>
</a>
</div>
<div class="base-col author-col">
<a href="" target="_blank" class="author-text">
Justin Kerr
<span class="text-star">1</span>
</a>
</div>
<div class="base-col author-col">
<a href="" target="_blank" class="author-text">
Richard Cheng
<span class="text-star">2</span>
</a>
</div>
<div class="base-col author-col">
<a href="" target="_blank" class="author-text">
Muhammad Zubair Irshad
<span class="text-star">2</span>
</a>
</div>
</div>
<div class="base-row author-row">
<div class="base-col author-col">
<a href="" target="_blank" class="author-text">
Ashwin Balakrishna
<span class="text-star">2</span>
</a>
</div>
<div class="base-col author-col">
<a href="" target="_blank" class="author-text">
Thomas Kollar
<span class="text-star">2</span>
</a>
</div>
<div class="base-col author-col">
<a href="https://goldberg.berkeley.edu/" target="_blank" class="author-text">
Ken Goldberg
<span class="text-star">1</span>
</a>
</div>
</div>
<div>
<!-- <h1 id="uc-berkeley">UC Berkeley</h1> -->
<h1 id="uc-berkeley"><span class="text-star">1</span> The AUTOLab at UC Berkeley</h1>
<h1 id="uc-berkeley"><span class="text-star">2</span> The Toyota Research Institute</h1>
<!-- <br /> -->
<div id="equal_contrib" style="margin-top: 7px;">
<span class="text-star">*</span>Denotes Equal Contribution
</div>
</div>
<div class="title-row">
<h2 class="subheader">IROS 2024 (Oral)</h1>
</div>
<div class="link-labels base-row">
<!-- TODO: Update arxiv link -->
<div class="base-col icon-col"><a href="https://arxiv.org/abs/2409.18108" target="_blank"
class="link-block"><img
src="https://uploads-ssl.webflow.com/51e0d73d83d06baa7a00000f/5cab99df4998decfbf9e218e_paper-01.png"
alt="paper"
sizes="(max-width: 479px) 12vw, (max-width: 767px) 7vw, (max-width: 991px) 41.8515625px, 56.6953125px"
srcset="https://uploads-ssl.webflow.com/51e0d73d83d06baa7a00000f/5cab99df4998decfbf9e218e_paper-01-p-500.png 500w, https://uploads-ssl.webflow.com/51e0d73d83d06baa7a00000f/5cab99df4998decfbf9e218e_paper-01.png 672w"
class="icon-img" /></a></div>
<!-- TODO: Update code link -->
<div class="base-col icon-col"><a href="https://github.com/BerkeleyAutomation/L3GS" class="link-block"><img
src="https://uploads-ssl.webflow.com/51e0d73d83d06baa7a00000f/5cae3b53b42ebb3dd4175a82_68747470733a2f2f7777772e69636f6e66696e6465722e636f6d2f646174612f69636f6e732f6f637469636f6e732f313032342f6d61726b2d6769746875622d3235362e706e67.png"
alt="paper" style="height: 60" class="icon-img github-img-icon" /></a></div>
</div>
<div class="link-labels base-row">
<div class="base-col icon-col">
<strong class="link-labels-text">Paper</strong>
</div>
<div class="base-col icon-col">
<strong class="link-labels-text">Code</strong>
</div>
</div>
<br>
<video controls autoplay loop width="800px">
<source src="LEGS_Kitchen_Queries.mp4" type="video/mp4" />
</video>
<br>
<br>
<video controls autoplay loop width="800px">
<source src="GroceryStoreQueries.mp4" type="video/mp4" />
</video>
<br>
<div class="base-row add-top-padding">
<h1 id="abstract">Overview</h1>
<p class="paragraph">
Building semantic 3D maps can be valuable for searching offices, warehouses, stores and homes for objects of interest. We present a multi-camera mapping system that incrementally builds a Language-Embedded Gaussian Splat (LEGS), a detailed 3D scene representation that encodes both appearance and semantics in a unified representation.
LEGS is trained online as the robot traverses its environment, enabling localization of open-vocabulary object queries.
We evaluate LEGS on three room-scale scenes where we query random objects in the scene to assess the system's ability to capture semantic meaning. We compare our system to LERF for these three scenes and find that while both systems have comparable object query success rates, LEGS trains over 3.5x faster than LERF.
Qualitative results suggest that multi-camera setup and incremental bundle adjustment boost visual reconstruction quality in constrained robot trajectories, and experimental results suggest LEGS can localize objects with up to 66% accuracy across three large indoor environments, and produce high fidelity Gaussian Splats in an online manner by integrating bundle adjustment updates.
</p>
</div>
<figure>
<img src="PSplash2.png" style="width: 100%;" alt="LEGS"/>
</figure>
<figure>
<img src="PSplash5.png" style="width: 100%;" alt="LEGS2"/>
<figcaption>Large-scale language-embedded Gaussian splatting setup. The
Gaussian splat 3D reconstruction was used to render a novel view of a
large-scale environment. Given open-vocabulary queries, LEGS can localize
the desired objects as seen with the heatmap activations.</figcaption>
</figure>
<div class="citation add-top-padding">
<h1 id="abstract"> Citation </h1>
<p> If you use this work or find it helpful, please consider citing: (bibtex) </p>
<pre id="codecell0">@article{yu2024language,
title={Language-Embedded Gaussian Splats (LEGS): Incrementally Building Room-Scale Representations with a Mobile Robot},
author={Yu, Justin and Hari, Kush and Srinivas, Kishore and El-Refai, Karim and Rashid, Adam
and Kim, Chung Min and Kerr, Justin and Cheng, Richard and Irshad, Muhammad Zubair
and Balakrishna, Ashwin and Kollar, Thomas and Goldberg, Ken},
journal={arXiv preprint arXiv:2409.18108},
year={2024}
} </pre>
</div>
<div>
<h1 id="abstract"> Acknowledgements </h1>
<p> This work was supported by the Toyota Research Institute (TRI)</p>
</div>
<div>
<h2 id="abstract"> For questions, please contact</h2>
<pre id="codecell0"> [email protected], [email protected]</pre>
</div>