-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
222 lines (190 loc) · 8.46 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
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<title>LifeQA: A Real-Life Dataset for Video Question Answering</title>
<meta name="description"
content="LifeQA is a benchmark dataset for Video Question Answering that focuses on day-to-day real-life situations.">
<meta name="keywords"
content="LifeQA, VideoQA, Video Question Answering, Computer Vision, Machine Learning, dataset, Natural Language Processing, Videos, YouTube, real life, research, LREC2020, LREC, Machine Learning, Deep Learning, NLP, PyTorch">
<meta name="author"
content="Santiago Castro, Mahmoud Azab, Jonathan C. Stroud, Cristina Noujaim, Ruoyao Wang, Jia Deng, and Rada Mihalcea">
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta property="og:type" content="website" />
<meta property="og:site_name" content="LifeQA: A Real-Life Dataset for Video Question Answering" />
<meta property="og:image" content="https://lit.eecs.umich.edu/lifeqa/img/example.png" />
<meta property="og:image:height" content="630" />
<meta property="og:image:width" content="1200" />
<meta property="og:title" content="LifeQA: A Real-Life Dataset for Video Question Answering" />
<meta property="og:description" content="LifeQA is a benchmark dataset for Video Question Answering that focuses on day-to-day real-life situations." />
<meta property="og:url" content="https://lit.eecs.umich.edu/lifeqa/" />
<meta name="twitter:card" content="summary_large_image" />
<meta name="twitter:site" content="@michigan_AI" />
<meta name="twitter:creator" content="@michigan_AI" />
<script async src="https://www.googletagmanager.com/gtag/js?id=UA-34392230-10"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'UA-34392230-10');
</script>
<link rel="stylesheet" type="text/css" href="main.css"/>
</head>
<body>
<div class="container">
<header>
<a href="https://www.tri.global/"><img id="tri" src="img/tri.png" alt="Toyota Research Institute logo"></a>
<a href="https://umich.edu/"><img id="um" src="img/um.png" alt="University of Michigan logo"></a>
<h1>LifeQA: A Real-Life Dataset for Video Question Answering</h1>
<ul id="quick-links">
<li><a href="http://www.lrec-conf.org/proceedings/lrec2020/pdf/2020.lrec-1.536.pdf">Paper</a></li>
<li><a href="https://github.com/mmazab/LifeQA">Data + Code</a></li>
<li><a href="https://www.aclweb.org/anthology/2020.lrec-1.536/">ACL Anthology page</a></li>
<li><a href="https://www.aclweb.org/anthology/2020.lrec-1.536.bib">BibTeX Citation</a></li>
</ul>
</header>
<section class="section-alt">
<div class="content">
<h2>Abstract</h2>
<p id="abstract">
We introduce <b>LifeQA</b>, a benchmark dataset for video question answering that focuses on day-to-day
real-life
situations. Current video question answering datasets consist of movies and TV shows. However, it is well-known
that these visual domains are not representative of our day-to-day lives. Movies and TV shows, for example,
benefit from professional camera movements, clean editing, crisp audio recordings, and scripted dialog between
professional actors. While these domains provide a large amount of data for training models, their properties
make
them unsuitable for testing real-life question answering systems. Our dataset, by contrast, consists of video
clips that represent only real-life scenarios. We collect 275 such video clips and over 2.3k multiple-choice
questions. In this paper, we analyze the challenging but realistic aspects of LifeQA, and we apply several
state-of-the-art video question answering models to provide benchmarks for future research.
</p>
</div>
</section>
<section>
<div class="content">
<a href="http://www.lrec-conf.org/proceedings/lrec2020/pdf/2020.lrec-1.536.pdf">
<ol id="thumbnails">
<li><img src="img/thumbs/0.png" alt="thumbnail, page 0"/></li>
<li><img src="img/thumbs/1.png" alt="thumbnail, page 1"/></li>
<li><img src="img/thumbs/2.png" alt="thumbnail, page 2"/></li>
<li><img src="img/thumbs/3.png" alt="thumbnail, page 3"/></li>
<li><img src="img/thumbs/4.png" alt="thumbnail, page 4"/></li>
<li><img src="img/thumbs/5.png" alt="thumbnail, page 5"/></li>
<li><img src="img/thumbs/6.png" alt="thumbnail, page 6"/></li>
</ol>
</a>
</div>
</section>
<section>
<div class="content">
<ol id="authors">
<li>
<a href="https://santi.uy">
<div class="author-img-container">
<img src="img/authors/santi.jpeg" alt="Santiago Castro profile picture">
</div>
Santiago Castro
</a>
</li>
<li>
<a href="https://web.eecs.umich.edu/~mazab/">
<div class="author-img-container">
<img src="img/authors/mazab.jpeg" alt="Mahmoud Azab profile picture">
</div>
Mahmoud Azab
</a>
</li>
<li>
<a href="https://www.jonathancstroud.com/">
<div class="author-img-container">
<img src="img/authors/jonathan.png" alt="Jonathan C. Stroud profile picture">
</div>
Jonathan C. Stroud
</a>
</li>
<li>
<div>
<div class="author-img-container">
<img src="img/authors/cristina.jpg" alt="Cristina Noujaim profile picture">
</div>
Cristina Noujaim
</div>
</li>
<li>
<a href="https://wsxzwps.github.io/">
<div class="author-img-container">
<img src="img/authors/ruoyao.jpg" alt="Ruoyao Wang profile picture">
</div>
Ruoyao Wang
</a>
</li>
<li>
<a href="https://www.cs.princeton.edu/~jiadeng/">
<div class="author-img-container">
<img src="img/authors/jia.jpg" alt="Jia Deng profile picture">
</div>
Jia Deng
</a>
</li>
<li>
<a href="https://web.eecs.umich.edu/~mihalcea/">
<div class="author-img-container">
<img src="img/authors/rada.jpg" alt="Rada Mihalcea profile picture">
</div>
Rada Mihalcea
</a>
</li>
</ol>
<p id="affiliation">
<a href="https://umich.edu/">
<img id="um-vertical" alt="University of Michigan" src="img/um-vertical.png">
</a>
</p>
</div>
</section>
<section class="section-alt">
<div class="content">
<h2>Downloads</h2>
<ul id="downloads">
<li><a href="http://www.lrec-conf.org/proceedings/lrec2020/pdf/2020.lrec-1.536.pdf">PDF Paper</a></li>
<li><a href="https://github.com/mmazab/LifeQA">Data + Code</a> (with instructions)</li>
<li>
<a href="https://drive.google.com/drive/folders/1sV1IYoC1oIgjHfSVkIJ-p8GA2hOwx4u1?usp=sharing">
Pre-extracted features</a>
(ResNet-152, C3D, and more)
</li>
<li>
<a href="https://github.com/mmazab/LifeQA/tree/master/data/lqa_trans">Manually transcribed captions</a>
</li>
</ul>
</div>
</section>
<!--section>
<div class="content">
<h2>Examples</h2>
<p></p>
</div>
</section-->
<footer>
<div class="content">
<h2>Acknowledgments</h2>
<p id="acknowledgments-text">
We are grateful to Aurelia Bunescu, <a href="https://dsouzadaniel.github.io/">Daniel D'Souza</a>,
<a href="https://haoopeng.github.io/">Penghao He</a>,
<a href="https://shubham14.github.io/">Shubham Dash</a>, and
<a href="http://www-personal.umich.edu/~ywchao/">Yu-Wei Chao</a> for their help with the collection and
annotation of the dataset. This material is based in part upon work supported by the
<a href="https://www.tri.global/">Toyota Research Institute ("TRI")</a>. Any opinions, findings, and conclusions
or recommendations expressed in this material are those of the authors and do not necessarily reflect the views
of TRI or any other Toyota entity.
</p>
<p>
Web page inspired by the
<a href="https://cs.stanford.edu/people/ranjaykrishna/densevid/">ActivityNet Captions web page</a>.
</p>
</div>
</footer>
</div>
</body>
</html>