-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathinterp-img-disc.html
148 lines (140 loc) · 5.91 KB
/
interp-img-disc.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
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no">
<link rel="stylesheet" href="https://stackpath.bootstrapcdn.com/bootstrap/4.1.1/css/bootstrap.min.css" integrity="sha384-WskhaSGFgHYWDcbwN70/dfYBj47jz9qbsMId/iRN3ewGhXQFZCSftd1LZCfmhktB" crossorigin="anonymous">
<link rel="stylesheet" type="text/css" href="assets/iid/css/iid.css">
</head>
<body>
<div class="container">
<div class="row">
<div class="col-12">
<br>
<h1 class="text-center">Interpretable Discovery in Large Image Data Sets</h1>
</div>
</div>
<br>
<div class="row">
<div class="col-lg-2"></div>
<div class="col-lg-4">
<table class="table table-borderless table-sm text-center">
<tbody>
<tr>
<td style="font-size:22px;"><a href="http://www.wkiri.com/">
Kiri Wagstaff
</a></td>
</tr>
<tr>
<td style="font-size:22px;">
Jet Propulsion Laboratory, <br>California Institute of Technology
</td>
</tr>
</tbody>
</table>
</div>
<div class="col-lg-4">
<table class="table table-borderless table-sm text-center">
<tbody>
<tr>
<td style="font-size:22px;"><a href="http://www.jakehlee.com/">
Jake Lee
</a></td>
</tr>
<tr>
<td style="font-size:22px;">
Columbia University
</td>
</tr>
</tbody>
</table>
</div>
<div class="col-lg-2"></div>
</div>
<div class="row" style="font-size:22px;">
<div class="col-lg-12 text-center">
<a href="https://sites.google.com/view/whi2018/home">WHI 2018</a> <a href="https://arxiv.org/pdf/1806.08340.pdf">[link to paper]</a>
</div>
</div>
<div class="row">
<div class="col-md-6">
<img src="assets/iid/img/teaser1.png" class="img-fluid" alt="ImageNet class discovery">
</div>
<div class="col-md-6">
<img src="assets/iid/img/teaser2.png" class="img-fluid" alt="Mars rover class discovery">
</div>
</div>
<hr>
<div>
<h1 class="text-center">Abstract</h1>
<br>
<p class="text-left">Automated detection of new, interesting, unusual, or anomalous images within large data sets has great value for applications from surveillance (e.g., airport security) to science (observations that don’t fit a given theory can lead to new discoveries). Many image data analysis systems are turning to convolutional neural networks (CNNs) to represent image content due to their success in achieving high classification accuracy rates. However, CNN representations are notoriously difficult for humans to interpret. We describe a new strategy that combines novelty detection with CNN image features to achieve rapid discovery with interpretable explanations of novel image content. We applied this technique to familiar images from ImageNet as well as to a scientific image collection from planetary science.</p>
</div>
<hr>
<div class="text-center">
<h1>Code & Data</h1>
<br>
<img src="assets/iid/img/diagram.png" class="img-fluid" alt="Interpretable image discovery diagram">
</div>
<br>
<dl class="row" style="font-size:22px;">
<dt class="col-sm-6 text-switch">DEMUD</dt>
<dd class="col-sm-6"><a href="https://github.com/wkiri/DEMUD">[GitHub repository]</a></dd>
<dt class="col-sm-6 text-switch">Supplemental scripts and ImageNet data</dt>
<dd class="col-sm-6"><a href="https://github.com/jakehlee/whi2018-interp-img-disc">[GitHub repository]</a></dd>
<dt class="col-sm-6 text-switch">Dosovitskiy and Brox visualization</dt>
<dd class="col-sm-6"><a href="https://lmb.informatik.uni-freiburg.de/resources/software.php">[Source]</a></dd>
<dt class="col-sm-6 text-switch">Mars image data set</dt>
<dd class="col-sm-6"><a href="http://doi.org/10.5281/zenodo.1049137">[Zenodo]</a></dd>
</dl>
<hr>
<div>
<h1 class="text-center">Paper</h1>
<br>
<div class="row">
<div class="col-lg-2"></div>
<div class="col-lg-4">
<div class="img-stack">
<a href="https://arxiv.org/pdf/1806.08340.pdf"><img src="assets/iid/pages/1.png" class="stack-1"></a>
<img src="assets/iid/pages/2.png" class="stack-2">
<img src="assets/iid/pages/3.png" class="stack-3">
</div>
</div>
<div class="col-lg-4" style="display: flex; align-items: center;">
<div class="text-left">
Kiri Wagstaff, Jake Lee.
<br>
<strong>Interpretable Discovery in Large Image Data Sets</strong>
<br>
In WHI 2018
<br>
<a href="https://arxiv.org/pdf/1806.08340.pdf">[link to paper]</a>
</div>
</div>
<div class="col-lg-2"></div>
</div>
</div>
<hr>
<div>
<h1 class="text-center">Presentation</h1>
<br>
<div class="row">
<div class="col-lg-2"></div>
<div class="col-lg-8 text-center">
<iframe width="560" height="315" src="https://www.youtube.com/embed/iByfVX_rnHY?rel=0" frameborder="0" allow="autoplay; encrypted-media" allowfullscreen></iframe>
</div>
<div class="col-lg-2"></div>
</div>
</div>
<hr>
<div>
<h1 class="text-center">Acknowledgements</h1>
<br>
<p>We thank the Planetary Data System Imaging Node for funding this project. Part of this research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration.</p>
</div>
</div>
<script src="https://code.jquery.com/jquery-3.3.1.slim.min.js" integrity="sha384-q8i/X+965DzO0rT7abK41JStQIAqVgRVzpbzo5smXKp4YfRvH+8abtTE1Pi6jizo" crossorigin="anonymous"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/popper.js/1.14.3/umd/popper.min.js" integrity="sha384-ZMP7rVo3mIykV+2+9J3UJ46jBk0WLaUAdn689aCwoqbBJiSnjAK/l8WvCWPIPm49" crossorigin="anonymous"></script>
<script src="https://stackpath.bootstrapcdn.com/bootstrap/4.1.1/js/bootstrap.min.js" integrity="sha384-smHYKdLADwkXOn1EmN1qk/HfnUcbVRZyYmZ4qpPea6sjB/pTJ0euyQp0Mk8ck+5T" crossorigin="anonymous"></script>
</body>
</html>