-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
executable file
·330 lines (329 loc) · 14.5 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
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
<!DOCTYPE html>
<html>
<head lang="en">
<meta charset="UTF-8">
<meta http-equiv="x-ua-compatible" content="ie=edge">
<title>Physically Grounded VLMs</title>
<meta name="description" content="Physically Grounded VLMs">
<meta name="viewport" content="width=device-width, initial-scale=1">
<link rel="stylesheet" href="https://maxcdn.bootstrapcdn.com/bootstrap/3.3.5/css/bootstrap.min.css">
<link rel="stylesheet" href="https://maxcdn.bootstrapcdn.com/font-awesome/4.4.0/css/font-awesome.min.css">
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/codemirror/5.8.0/codemirror.min.css">
<link rel="stylesheet" href="css/app.css">
<link rel="stylesheet" href="css/bootstrap.min.css">
<script src="https://ajax.googleapis.com/ajax/libs/jquery/1.11.3/jquery.min.js"></script>
<script src="https://maxcdn.bootstrapcdn.com/bootstrap/3.3.5/js/bootstrap.min.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/codemirror/5.8.0/codemirror.min.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/clipboard.js/1.5.3/clipboard.min.js"></script>
<script src="js/app.js"></script>
<!-- Google tag (gtag.js) -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-52J0PM8XKV"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'G-52J0PM8XKV');
</script>
<style>
.nav-pills {
position: relative;
display: inline;
}
.imtip {
position: absolute;
top: 0;
left: 0;
}
</style>
</head>
<body>
<div class="container" id="main">
<div class="row">
<h2 class="col-md-12 text-center">
<strong><font size="+6">Physically Grounded Vision-Language Models</font></strong> </br> for Robotic Manipulation </br>
</h2>
</div>
<div class="row">
<div class="col-md-12 text-center">
<ul class="list-inline">
<li>Jensen Gao</li>
<li>Bidipta Sarkar</li>
<li>Fei Xia</li>
<li>Ted Xiao</li>
<li>Jiajun Wu</li>
<li>Brian Ichter</li>
<li>Anirudha Majumdar</li>
<li>Dorsa Sadigh</li>
<br>
<a href="https://stanford.edu">
<image src="img/stanford_logo.png" height="55px">
</a>
<a href="http://g.co/robotics">
<image src="img/rng-logo.png" height="47px">
</a>
<br>
<font size="+2"><li>ICRA 2024</li></font>
</ul>
</div>
</div>
<div class="row">
<div class="col-md-6 col-md-offset-3 text-center">
<ul class="nav nav-pills nav-justified">
<li>
<a href="https://arxiv.org/abs/2309.02561">
<image src="img/arxiv.jpg" height="60px">
<h4><strong>Paper</strong></h4>
</a>
</li>
<li>
<a href="appendix/appendix.pdf">
<image src="img/appendix.jpg" height="60px">
<h4><strong>Appendix</strong></h4>
</a>
</li>
<li>
<a href="https://drive.google.com/file/d/17gbzrJSs8YjVafIrX4omR_rx6qLgXjUd/view?usp=sharing">
<image src="img/youtube_icon.png" height="60px">
<h4><strong>Video</strong></h4>
</a>
</li>
<!-- https://iconduck.com/icons/193266/dataset -->
<li>
<a href="https://drive.google.com/file/d/1ThZ7p_5BnMboK_QE13m1fPKa4WGdRcfC/view?usp=sharing">
<image src="img/dataset.svg" height="60px">
<h4><strong>Dataset</strong></h4>
</a>
</li>
<li>
<a href="https://huggingface.co/bidiptas/PG-InstructBLIP">
<image src="img/huggingface.svg" height="60px">
<h4><strong>Model</strong></h4>
</a>
</li>
</ul>
</div>
</div>
<div class="row">
<div class="col-md-8 col-md-offset-2">
<p style="text-align:center;">
<div style="position:relative;padding-top:56.25%;">
<iframe src="https://drive.google.com/file/d/17gbzrJSs8YjVafIrX4omR_rx6qLgXjUd/preview" width="560" height="315" allowfullscreen style="position:absolute;top:0;left:0;width:100%;height:100%;"></iframe>
</div>
</p>
<h3>
Abstract
</h3>
<p class="text-justify">
Recent advances in vision-language models (VLMs) have led to improved performance on tasks such as visual question answering and image captioning. Consequently, these models are now well-positioned to reason about the physical world, particularly within domains such as robotic manipulation. However, current VLMs are limited in their understanding of the physical concepts (e.g., material, fragility) of common objects, which restricts their usefulness for robotic manipulation tasks that involve interaction and physical reasoning about such objects. To address this limitation, we propose PhysObjects, an object-centric dataset of 39.6K crowd-sourced and 417K automated physical concept annotations of common household objects. We demonstrate that fine-tuning a VLM on PhysObjects improves its understanding of physical object concepts, including generalization to held-out concepts, by capturing human priors of these concepts from visual appearance. We incorporate this physically grounded VLM in an interactive framework with a large language model-based robotic planner, and show improved planning performance on tasks that require reasoning about physical object concepts, compared to baselines that do not leverage physically grounded VLMs. We additionally illustrate the benefits of our physically grounded VLM on a real robot, where it improves task success rates.
</p>
<h3>
PhysObjects Dataset
</h3>
<p class="text-justify">
To benchmark and improve VLMs for object-centric physical reasoning, we compiled the PhysObjects dataset, which contains 39.6K crowd-sourced and 417K automated physical concept annotations. The source of our images is the <a href="https://ai.meta.com/datasets/egoobjects-dataset/">EgoObjects dataset</a>.
We collected annotations for eight physical concepts, listed in the table below. We chose these concepts based on prior work and what we believe to be userful for robotic manipulation. However, we do not consider concepts that would be difficult for humans to estimate from only images, such as friction.
<table id="tbl">
<tr>
<th> Concept </th>
<th> Description </th>
</tr>
<tr>
<td> Mass </td>
<td> How heavy an object is </td>
</tr>
<tr>
<td> Fragility </td>
<td> How easily an object can be broken/damaged </td>
</tr>
<tr>
<td> Deformability </td>
<td> How easily an object can change shape without breaking </td>
</tr>
<tr>
<td> Material </td>
<td> What an object is primarily made of </td>
</tr>
<tr>
<td> Transparency </td>
<td> How much can be seen through an object </td>
</tr>
<tr>
<td> Contents </td>
<td> What is inside a container </td>
</tr>
<tr>
<td> Can Contain Liquid </td>
<td> If a container can be used to easily carry liquid </td>
</tr>
<tr>
<td> Is Sealed </td>
<td> If a container will not spill if rotated </td>
</tr>
<tr>
<td> Density <i>(held-out)</i> </td>
<td> How much mass per unit of volume of an object </td>
</tr>
<tr>
<td> Liquid Capacity <i>(held-out)</i> </td>
<td> How much liquid a container can contain </td>
</tr>
</table>
</p>
</div>
</div>
<div class="row">
<div class="col-md-8 col-md-offset-2">
<br>
<h3>
Real Scene Planning Evaluation
</h3>
<p class="text-justify">
Click on a scene image to go to the planning results for a task in that scene.
<div class="irow">
<div class="column">
<h4>
Scene 1: Countertop
</h4>
<p style="text-align:center;">
<a href="scene1.html">
<img src="img/eval2_scene1.png" class="img-responsive">
</a>
</p>
</div>
<div class="column">
<h4>
Scene 2: Art Table
</h4>
<p style="text-align:center;">
<a href="scene2.html">
<img src="img/eval2_scene2.png" class="img-responsive">
</a>
</p>
</div>
</div>
<div class="irow">
<div class="column">
<h4>
Scene 3: Floor
</h4>
<p style="text-align:center;">
<a href="scene3.html">
<img src="img/eval2_scene3.png" class="img-responsive">
</a>
</p>
<h4>
Scene 5: Kitchen B
</h4>
<p style="text-align:center;">
<a href="scene5.html">
<img src="img/eval2_scene5.png" class="img-responsive">
</a>
</p>
<h4>
Scene 7: Living Room
</h4>
<p style="text-align:center;">
<a href="scene7.html">
<img src="img/eval2_scene7.png" class="img-responsive">
</a>
</p>
</div>
<div class="column">
<h4>
Scene 4: Kitchen A
</h4>
<p style="text-align:center;">
<a href="scene4.html">
<img src="img/eval2_scene4.png" class="img-responsive">
</a>
</p>
<h4>
Scene 6: Salad Bar
</h4>
<p style="text-align:center;">
<a href="scene6.html">
<img src="img/eval2_scene6.png" class="img-responsive">
</a>
</p>
<h4>
Scene 8: Shelf
</h4>
<p style="text-align:center;">
<a href="scene8.html">
<img src="img/eval2_scene8.png" class="img-responsive">
</a>
</p>
</div>
</div>
</div>
<div class="row">
<div class="col-md-8 col-md-offset-2">
<br>
<h3>
Real Robot Evaluation
</h3>
<p class="text-justify">
Click on a scene image to go to videos of all tasks for that scene.
<div class="irow">
<div class="column">
<h4>
Robot Scene 1
</h4>
<p style="text-align:center;">
<a href="RS1/index.html">
<img src="img/RealScene1.png" class="img-responsive">
</a>
</p>
</div>
<div class="column">
<h4>
Robot Scene 2
</h4>
<p style="text-align:center;">
<a href="RS2/index.html">
<img src="img/RealScene2.png" class="img-responsive">
</a>
</p>
</div>
</div>
</div>
</div>
<div class="row">
<div class="col-md-8 col-md-offset-2">
<h3>
Citation
</h3>
<div class="form-group col-md-10 col-md-offset-1">
<textarea id="bibtex" class="form-control" readonly>@inproceedings{pgvlm2024,
title={Physically Grounded Vision-Language Models for Robotic Manipulation},
author={Jensen Gao and Bidipta Sarkar and Fei Xia and Ted Xiao and Jiajun Wu
and Brian Ichter and Anirudha Majumdar and Dorsa Sadigh},
booktitle={IEEE International Conference on Robotics and Automation (ICRA)},
year={2024},
organization={IEEE}
} </textarea>
</div>
</div>
</div>
<div class="row">
<div id="open-source" class="col-md-8 col-md-offset-2">
<h3>
Open Source
</h3>
We open source the PG-InstructBLIP model <a href="https://huggingface.co/bidiptas/PG-InstructBLIP">[here]</a>.
We also open source the dataset <a href="https://drive.google.com/file/d/1ThZ7p_5BnMboK_QE13m1fPKa4WGdRcfC/view?usp=sharing">[here]</a>.
<p style="text-align:center;">
</p>
</div>
</div>
<div class="row">
<div class="col-md-8 col-md-offset-2">
<p class="text-justify">
<br><br>
The website template was borrowed from <a href="http://jonbarron.info/">Jon Barron</a> and <a href="https://robotics-transformer.github.io/">RT-1</a>
</p>
</div>
</div>
</div>
</body>
</html>