-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathscene8.html
175 lines (154 loc) · 6.89 KB
/
scene8.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
<!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">
<!-- <base href="/"> -->
<!-- <link rel="apple-touch-icon" href="apple-touch-icon.png"> -->
<!-- <link rel="icon" type="image/png" href="img/seal_icon.png"> -->
<!-- Place favicon.ico in the root directory -->
<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="topnav">
<a href="index.html">Home</a>
<div class="rightnav">
<a href="scene1.html">Scene 1: Countertop</a>
<a href="scene2.html">Scene 2: Art Table</a>
<a href="scene3.html">Scene 3: Floor</a>
<a href="scene4.html">Scene 4: Kitchen A</a>
<a href="scene5.html">Scene 5: Kitchen B</a>
<a href="scene6.html">Scene 6: Salad Bar</a>
<a href="scene7.html">Scene 7: Living Room</a>
<a href="scene8.html" class="active">Scene 8: Shelf</a>
<a href="RS1/index.html">Robot Scene 1</a>
<a href="RS2/index.html">Robot Scene 2</a>
</div>
</div>
<div class="container" id="main">
<div class="row">
<div class="col-md-8 col-md-offset-2">
<br>
<h1>
Scene 8: Shelf
</h1>
<div class="irow">
<div class="strict_column">
<img src="img/eval2_scene8.png">
</div>
<div class="strict_column">
<p> A: box [binder] </p>
<p> B: bottle [large plastic tub] </p>
<p> C: bottle [plastic bottle] </p>
<p> D: box [algorithms textbook] </p>
<p> E: pitcher (container) [blue metal cup] </p>
<p> F: water glass [small glass cup] </p>
<p style="color: #ff00ff"> G: headphones [phone cable] </p>
<p> H: dumbbell [power brick] </p>
<p style="color: #9900ff"> I: adhesive tape [ruler] </p>
</div>
</div>
<br>
<p class="text-justify">
Here we provide a scene from our real scene planning evaluation, and one example task for it. We provide the object detections and labels from OWL-ViT, which is what our planner has access to, but we provide more precise labels in square brackets. We provide color-coded bounding boxes for the object detections that are relevant for the example task. For each version of the LLM planner, we provide its output, including possible interaction with a VLM, and the generated plan. Outputs generated by the VLMs are in <span style="color: #ff9900">orange</span>, and we mark correct and incorrect plans with <span style="color: #38761d">green</span> and <span style="color: #ff0000">red</span> respectively.
</p>
<h2><b>Task 48: Bring me the most bendable object.</b></h2>
<p class="text-justify">
In this task, we want to determine if the VLM understands which objects are bendable (which is a synonym for deformable). The best object in this scene is <span style="color: #ff00ff"> G </span>, which is a cable to charge devices.
</p>
<h3>Generated Plans</h3>
<div class="irow">
<div class="tcolumn">
<h3> No VLM </h3>
<p> Thought: The most bendable object among the given objects is the adhesive tape. </p>
<p> Plan: </p>
<p style="color: #ff0000">
1. Go to object I <br>
2. Pick up object I <br>
3. Bring to human object I <br>
4. Done</p>
</div>
<div class="tcolumn">
<h3> InstructBLIP </h3>
<p> Thought: I can ask which object is bendable out of all the objects, and I can find the one with the largest Yes value. </p>
<p> Question about [A, B, C, D, E, F, G, H, I]: Is this object bendable? </p>
<p style="color: #ff9900"> Answer <br>
A: Yes (0.53), No (0.42), Unknown (0.03) <br>
B: No (0.65), Yes (0.29), Unknown (0.03) <br>
C: No (0.72), Yes (0.23), Unknown (0.03) <br>
D: No (0.6), Yes (0.36), Unknown (0.02) <br>
E: No (0.85), Yes (0.11), Unknown (0.02) <br>
F: No (0.62), Yes (0.29), Unknown (0.05) <br>
G: Yes (0.63), No (0.33), Unknown (0.02) <br>
H: No (0.51), Yes (0.42), Unknown (0.04) <br>
I: No (0.53), Yes (0.44), Unknown (0.01)
</p>
<p> Thought: The most bendable object is the one with the highest Yes value or the lowest No value (if none are Yes). The most bendable object is G (headphones) with the highest Yes value. </p>
<p> Plan: </p>
<p style="color: #38761d">
1. Go to object G <br>
2. Pick up object G <br>
3. Bring to human object G <br>
4. Done
</p>
</div>
<div class="tcolumn">
<h3> PG-InstructBLIP (ours) </h3>
<p> Thought: I can ask which object is bendable out of all the objects, and I can find the one with the largest Yes value. </p>
<p> Question about [A, B, C, D, E, F, G, H, I]: Is this object bendable? </p>
<p style="color: #ff9900"> Answer <br>
A: Yes (0.6), No (0.38), Unknown (0.01) <br>
B: No (0.7), Yes (0.26), Unknown (0.02) <br>
C: No (0.62), Yes (0.35), Unknown (0.02) <br>
D: No (0.52), Yes (0.45), Unknown (0.02) <br>
E: No (0.95), Yes (0.03), Unknown (0.0) <br>
F: No (0.86), Yes (0.11), Unknown (0.01) <br>
G: Yes (0.98), No (0.02) <br>
H: No (0.73), Yes (0.24), Unknown (0.02) <br>
I: Yes (0.56), No (0.42), Unknown (0.01) <br>
</p>
<p> Thought: The most bendable object is the one with the highest Yes value. The highest Yes value is for object G (headphones). </p>
<p> Plan: </p>
<p style="color: #38761d">
1. Go to object G <br>
2. Pick up object G <br>
3. Bring to human object G <br>
4. Done
</p>
</div>
</div>
</div>
</div>
</body>
</html>