-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathindex.html
231 lines (217 loc) · 10.3 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
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>ASGrasp</title>
<!-- Bootstrap -->
<link rel="preconnect" href="https://rsms.me/">
<link rel="stylesheet" href="https://rsms.me/inter/inter.css">
<link href="css/bootstrap-4.4.1.css" rel="stylesheet">
<link href="css/main.css" rel="stylesheet">
<!-- MathJax Script -->
<script src="https://polyfill.io/v3/polyfill.min.js?features=es6"></script>
<script id="MathJax-script" async src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js"></script>
<style>
body {
background: rgb(255, 255, 255) no-repeat fixed top left;
font-family: "Inter", 'Open Sans', sans-serif;
}
</style>
</head>
<!-- cover -->
<section>
<div class="jumbotron text-center mt-0">
<div class="container-fluid">
<div class="row">
<div class="col">
<h2 style="font-size:30px;">ASGrasp: Generalizable Transparent Object Reconstruction
and 6-DoF Grasp Detection from RGB-D Active Stereo Camera</h2>
<h4 style="color:#6e6e6e;"> ICRA 2024 </h4>
<hr>
<h6>
Jun Shi<sup>1*</sup>
Yong A<sup>1</sup>
Yixiang Jin<sup>1</sup>
Dingzhe Li<sup>1</sup>
Haoyu Niu<sup>1</sup>
Zhezhu Jin<sup>1</sup>
<a href="https://hughw19.github.io/" target="_blank">He Wang</a><sup>2,3,4†</sup>
<br>
<br>
<p>
<sup>1</sup>Samsung R&D Institute China-Beijing
<sup>2</sup>CFCS, Peking University
<sup>3</sup>Gallbot
<sup>4</sup>Beijing Academy of Artificial Intelligence (BAAI)
<br>
</p>
<p> <sup>†</sup> corresponding author
<br>
</p>
<div class="row justify-content-center">
<div class="column">
<p class="mb-5">
<a class="btn btn-large btn-light" href="https://arxiv.org/abs/2405.05648" role="button" target="_blank">
<!-- <a class="btn btn-large btn-light" href="https://arxiv.org/abs/2211.05272" role="button" target="_blank"> -->
<i class="fa fa-file"></i> Paper </a> </p>
</div>
<div class="column">
<p class="mb-5"><a class="btn btn-large btn-light" href="https://github.com/jun7-shi/ASGrasp" role="button" target="_blank">
<!-- <p class="mb-5"><a class="btn btn-large btn-light" href="https://github.com/PKU-EPIC/GAPartNet" role="button" target="_blank"> -->
<i class="fa fa-github-alt"></i> Code </a> </p>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- teaser -->
<section>
<div class="container">
<div class="row">
<div class="col-12 text-center">
<hr style="margin-top:0px">
<div class="row justify-content-center" style="align-items:center; display:flex;">
<img src="images/Fig1.PNG" alt="input" class="img-responsive graph" width="95%"/>
<br>
</div>
<p class="text-justify">
We propose ASGrasp, a 6-DoF <b>Grasp</b> detection network that uses an RGB-D <b>A</b>ctive <b>S</b>tereo camera,
to tackle the problem of grasping transparent and specular objects.
</p>
</div>
</div>
</div>
</section>
<br>
<!-- abstract -->
<section>
<div class="container">
<div class="row">
<div class="col-12">
<h2><strong>Abstract</strong></h2>
<hr style="margin-top:0px">
<p class="text-justify">
In this paper, we tackle the problem of grasping transparent and specular objects.
This issue holds importance, yet it remains unsolved within the field of robotics due to failure of recover their accurate geometry by depth cameras.
For the first time, we propose ASGrasp, a 6-DoF grasp detection network that uses an RGB-D active stereo camera.
ASGrasp utilizes a two-layer learning-based stereo network for the purpose of transparent object reconstruction, enabling material-agnostic object grasping in cluttered environments.
In contrast to existing RGB-D based grasp detection methods, which heavily depend on depth restoration networks and the quality of depth maps generated by depth cameras,
our system distinguishes itself by its ability to directly utilize raw IR and RGB images for transparent object geometry reconstruction.
We create an extensive synthetic dataset through domain randomization, which is based on GraspNet-1Billion.
Our experiments demonstrate that ASGrasp can achieve over 90% success rate for generalizable transparent object grasping in both simulation and the real via seamless sim-to-real transfer. Our method significantly outperforms SOTA networks and even surpasses the performance upper bound set by perfect visible point cloud inputs.
</p>
</div>
</div>
</div>
</section>
<br>
<!-- Methods -->
<section>
<div class="container">
<div class="row">
<div class="col-12">
<h2><strong>Methods</strong></h2>
<hr style="margin-top:0px">
<h3 style="margin-top:20px; margin-bottom:20px; color:#717980"><b>Full pipeline</b></h3>
<div class="row justify-content-center" style="align-items:center; display:flex;">
<img src="images/Framework.PNG" alt="input" class="img-responsive graph" width="95%"/>
</div>
<p class="text-justify">
<b>The architecture of our proposed approach involves the extraction of features
\(F_c\), \({F_{ir}^{l}}\), and \({F_{ir}^{r}}\) from the RGB image
\({I_c}\), the left IR image \({I_{ir}^{l}}\), and the right IR image \({I_{ir}^{r}}\).
\({F_{ir}^{l}}\) and \({F_{ir}^{r}}\) are used to construct a correlation pyramid.
This correlation pyramid, along with \({F_c}\), is fed into a GRU network for the prediction of a complete point cloud,
which includes the first-layer depth (visible point cloud) and the second-layer depth (invisible point cloud).
Subsequently, GSNet generates grasp poses based on the complete point cloud.
</p>
<br>
</div>
</div>
</div>
</div>
</section>
<br>
<br>
<!-- Results -->
<section>
<div class="container">
<div class="row">
<div class="col-12">
<h2><strong>Results</strong></h2>
<hr style="margin-top:0px">
<!-- <h3 style="margin-top:20px; margin-bottom:20px; color:#717980"><b>Cross-Category Part Segmentation</b></h3> -->
We visualize a qualitative comparison of the predictions for both layers.
The simulated raw point cloud contains missing and wrong points.
Our method performs much better in both the first and second layers, yielding a high-quality complete point cloud reconstruction.
<p class="text-justify">
<div class="row justify-content-center" style="align-items:center; display:flex;">
<img src="images/depth_comparison.PNG" alt="input" class="img-responsive graph" width="95%"/>
</div>
</p>
We tested our method in real-world experiments, showing sim2real results.
The test cases demonstrate significant potential in our methods for grasping transparent and specular objects.
<p class="text-justify">
<div class="row justify-content-center" style="align-items:center; display:flex;">
<img src="images/ASGrasp_mixscene.gif" alt="input" class="img-responsive graph" width="95%"/>
</div>
</p>
<p class="text-justify">
<div class="row justify-content-center" style="align-items:center; display:flex;">
<img src="images/ASGrasp_transparent.gif" alt="input" class="img-responsive graph" width="95%"/>
</div>
</p>
<p class="text-justify">
<div class="row justify-content-center" style="align-items:center; display:flex;">
<img src="images/ASGrasp_extreme.gif" alt="input" class="img-responsive graph" width="95%"/>
</div>
</p>
</div>
</div>
</div>
</div>
</section>
<br>
<!-- citing -->
<div class="container">
<div class="row ">
<div class="col-12">
<h2><strong>Citation</strong></h2>
<hr style="margin-top:0px">
<!-- <pre style="background-color: #e9eeef;padding: 1.25em 1.5em"> -->
<pre style="background-color: #e9eeef;padding: 0 1.5em">
<code>
@article{shi2024gsnet,
title={ASGrasp: Generalizable Transparent Object Reconstruction and 6-DoF Grasp Detection from RGB-D Active Stereo Camera},
author={Jun Shi, Yong A, Yixiang Jin, Dingzhe Li, Haoyu Niu, Zhezhu Jin, He Wang},
journal={arXiv preprint arXiv:2405.05648},
year={2024}
}
</code>
</pre>
</div>
</div>
</div>
<br>
<!-- Contact -->
<div class="container">
<div class="row ">
<div class="col-12">
<h2><strong>Contact</strong></h2>
<hr style="margin-top:0px">
<p>If you have any questions, please feel free to contact us:
<ul>
<li><b>Jun Shi</b>: jun7.shi<span style="display:none">Prevent spamming</span>@<span style="display:none">Prevent spamming</span>samsung.com </li>
<li><b>He Wang</b>: hewang<span style="display:none">Prevent spamming</span>@<span style="display:none">Prevent spamming</span>pku.edu.cn </li>
</ul>
</p>
</pre>
</div>
</div>
</div>
</body>
</html>