-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathindex.html
408 lines (361 loc) · 15.3 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
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
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
<html lang="en" class="">
<head>
<!-- Required meta tags -->
<title>ANetQA</title>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no">
<meta name="keywords" content="video, reasoning, physics, deep learning, computer vision, machine learning">
<meta name="description"
content="ANetQA: A Large-scale Benchmark for Fine-grained Compositional Reasoning over Untrimmed Videos">
<!-- Bootstrap CSS -->
<link rel="stylesheet" href="https://maxcdn.bootstrapcdn.com/bootstrap/3.3.7/css/bootstrap.min.css">
<link rel="stylesheet" href="https://maxcdn.bootstrapcdn.com/bootstrap/3.3.7/css/bootstrap-theme.min.css">
<script type="text/javascript" async="" src="https://www.google-analytics.com/analytics.js"></script>
<script src="https://code.jquery.com/jquery-3.2.1.slim.min.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/popper.js/1.11.0/umd/popper.min.js"></script>
<script src="https://maxcdn.bootstrapcdn.com/bootstrap/3.3.7/js/bootstrap.min.js"></script>
<script async="" src="https://www.googletagmanager.com/gtag/js?id=UA-81724582-4"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag() { dataLayer.push(arguments) };
gtag('js', new Date());
gtag('config', 'UA-81724582-4');
</script>
<style>
body {
font-size: 16px
}
.navbar-fixed-top {
min-height: 60px;
}
.navbar-nav {
margin-left: -20px;
}
.navbar-nav>li>a {
padding-top: 10px;
padding-bottom: 0px;
line-height: 60px;
font-size: 22px;
color: gray;
}
.navbar-nav>li>a:active {
color: black;
}
.navbar-nav>li>a:hover {
color: black;
-webkit-tap-highlight-color: rgba(0, 0, 0, 0);
-webkit-tap-highlight-color: transparent;
outline: none;
background: none;
text-decoration: none;
}
.col-md-13 {
display: inline-block;
text-align: justify;
}
.anet_logo {
margin-top: -30px;
margin-right: 30px;
float: left;
}
.logo {
margin-top: -80px;
margin-right: 30px;
float: right;
}
.logo1 {
margin-right: 30px;
float: right;
}
.logo2 {
margin-right: 30px;
float: right;
}
.logo3 {
margin-top: 20px;
margin-right: 30px;
float: right;
}
.img1 {
float: left
}
.img2 {
float: left
}
.sg {
margin-right: 20%;
float: right
}
.qa {
float: left
}
</style>
<!-- Custom styles for this template -->
<link href="jumbotron.css" rel="stylesheet">
<style>
hcfy-result.__hcfy__result__loaded__.__hcfy__result__both__ {
border: 1px dotted
}
</style>
</head>
<body data-gr-c-s-loaded="true">
<nav class="navbar-fixed-top" style="background-color: rgb(255,255,255)">
<!-- <a class="navbar-brand" href="#" style="font-size: 25px; color:white;line-height: 100%;">ANetQA</a> -->
<div class="anet_logo">
<img src="logo.jpg" height="125" width="125">
</div>
<div class="collapse navbar-collapse" id="navbarsExampleDefault">
<ul class="nav navbar-nav mr-auto">
<li><a href="#Paper" style="font-size: 20px">Paper</a></li>
<li><a href="#Dataset" style="font-size: 20px">Dataset</a></li>
<li><a href="#Code" style="font-size: 20px">Code</a></li>
<li><a href="#Eval" style="font-size: 20px">Evaluation</a></li>
<li><a href="#Licence" style="font-size: 20px">Licence</a></li>
</ul>
</div>
<div class="logo">
<div class="logo3">
<a href="https://research.lenovo.com/webapp/view/index.html"><img class="img-responsive img-rounded"
src="lenovo_logo.png" height="30" width="150"></a>
</div>
<div class="logo1">
<a href="https://www.zju.edu.cn/"><img class="img-responsive img-rounded" src="zju_logo.png" height="75"
width="75"></a>
</div>
<div class="logo2">
<a href="https://www.hdu.edu.cn"><img class="img-responsive img-rounded" src="hdu_logo.jpg" height="75"
width="75"></a>
</div>
</div>
</nav>
<main role="main">
<div class="container" style="padding-top: 80px; font-size: 20px">
<div align="center">
<h1 class="text-center" aligh="center">
ANetQA: A Large-scale Benchmark for Fine-grained
<br>
Compositional Reasoning over Untrimmed Videos
</h1><br>
<b>Zhou Yu<sup>1</sup></b>
<b>Lixiang Zheng<sup>1</sup></b>
<b>Zhou Zhao<sup>2</sup></b>
<b>Fei Wu<sup>2</sup></b>
<b>Jianping Fan<sup>1,3</sup></b>
<b>Kui Ren<sup>4</sup></b></a>
<b>Jun Yu<sup>1*</sup></b></a>
<br><br>
<h5><sup>1</sup>School of Computer Science, Hangzhou Dianzi University, China</h5>
<h5><sup>2</sup>Colledge of Computer Science and Technology, Zhejiang University, China</h5>
<h5><sup>3</sup>AI Lab at Lenovo Research, China</h5>
<h5><sup>4</sup>School of Cyber Science and Technology, Zhejiang University, China</h5>
<h5><sup>*</sup>Corresponding author</h5>
</div>
</div>
<br><br>
<div class="container">
<h2 id="RFSleep" style="padding-top: 80px; margin-top: -80px;">Abstract</h2>
<hr>
<div class="row">
<!-- <div class="col-md-10 col-md-offset-1"> -->
<div class="col-md-13" style="margin-left: 20px;margin-right: 20px;">
Building benchmarks to systemically analyze different
capabilities of video question answering (VideoQA) models
is challenging yet crucial. Existing benchmarks often
use non-compositional simple questions and suffer from
language biases, making it difficult to diagnose model
weaknesses incisively. A recent benchmark AGQA poses
a promising paradigm to generate QA pairs automatically
from pre-annotated scene graphs, enabling it to measure
diverse reasoning abilities with granular control. However,
its questions have limitations in reasoning about the finegrained
semantics in videos as such information is absent
in its scene graphs. To this end, we present ANetQA, a
large-scale benchmark that supports fine-grained compositional
reasoning over the challenging untrimmed videos
from ActivityNet. Similar to AGQA, the QA pairs
in ANetQA are automatically generated from annotated
video scene graphs. The fine-grained properties of ANetQA
are reflected in the following: (i) untrimmed videos with
fine-grained semantics; (ii) spatio-temporal scene graphs
with fine-grained taxonomies; and (iii) diverse questions
generated from fine-grained templates. ANetQA attains 1.4
billion unbalanced and 13.4 million balanced QA pairs,
which is an order of magnitude larger than AGQA with
a similar number of videos. Comprehensive experiments
are performed for state-of-the-art methods. The best model
achieves 44.5% accuracy while human performance tops
out at 84.5%, leaving sufficient room for improvements.</div>
<div style="margin-top: 20px">
<!-- <div class="col-md-10 col-md-offset-1"> -->
<div style="margin-left: 20px">
<img class="img-responsive img-rounded img1" src="anetqaex.jpg" alt="">
</div>
</div>
</div>
<br><br>
<div class="container">
<h2 id="Paper" style="padding-top: 80px; margin-top: -80px;">Paper</h2>
<hr>
<div class="row">
<div class="col-md-9">
<a href="https://arxiv.org/abs/2305.02519"><b>
ANetQA: A Large-scale Benchmark for Fine-grained Compositional Reasoning over Untrimmed Videos
</b><br></a>
Zhou Yu, Lixiang Zheng, Zhou Zhao, Fei Wu, Jianping Fan, Kui Ren, Jun Yu <br>
<i> CVPR, 2023</i><br>
<a href="https://arxiv.org/abs/2305.02519">[PDF]</a>
</div>
</div>
</div><br><br>
<!-- -->
<div class="container">
<h2 id="Dataset" style="padding-top: 80px; margin-top: -80px;">Dataset</h2>
<hr>
<h3>Videos</h3>
<div class="row">
<div class="col-md-9">
<li>Raw videos from <a href="http://activity-net.org/download.html" class="download-link">ActivityNet
v1.3</a></li>
</div>
</div>
<div class="row">
<div class="col-md-9">
<li><a
href="https://awma1-my.sharepoint.com/:u:/g/personal/yuz_l0_tn/EYIaBMbntepBt2tiG7USPO8Byi3ap-MkltQNdtUh9vZ2_w?download=1"
class="download-link">Meta information</a> of all videos.</li>
</div>
</div>
<hr>
<div class="sg">
<h3>Scene Graphs</h3>
<div class="row">
<div class="col-md-12">
<li><a
href="https://awma1-my.sharepoint.com/:u:/g/personal/yuz_l0_tn/ESVPcVYJlXZAlD5IkD_9y80BYcsfpC7Gp9LJGfXJYqreSw?download=1"
class="download-link">Train scene graphs</a> from 9,155 videos.</li>
</div>
</div>
<div class="row">
<div class="col-md-12">
<li><a
href="https://awma1-my.sharepoint.com/:u:/g/personal/yuz_l0_tn/Ea2UrDyFRTNNisb3BtJj6eAB5uoI1A7ewXam7OesPpFqzg?download=1"
class="download-link">Val scene graphs</a> from 1,185 videos.</li>
</div>
</div>
<div class="row">
<div class="col-md-12">
<li><a
href="https://awma1-my.sharepoint.com/:u:/g/personal/yuz_l0_tn/ESY5fUliVkZOo90ys6MdOGEBmKL8cpkV9kiwTKW0sAvsIQ?download=1"
class="download-link">Meta information</a> of all scene graphs.</li>
</div>
</div>
</div>
<div class="qa">
<h3>Question-Answer Pairs</h3>
<div class="row">
<div class="col-md-12">
<li><a
href="https://awma1-my.sharepoint.com/:u:/g/personal/yuz_l0_tn/EbOk0tgkpZlIqwsW1yVq7PgB2jJUx0x0eCv5iu73hl--uQ?download=1"
class="download-link">Train QA pairs</a> (10,456,011 samples)</li>
</div>
</div>
<div class="row">
<div class="col-md-12">
<li><a
href="https://awma1-my.sharepoint.com/:u:/g/personal/yuz_l0_tn/EckH2khIKF9PqX4lWwN2uJ0Bn_jA1Qvzv08Ny9jxEuzfWw?download=1"
class="download-link">Val QA pairs</a> (1,474,723 samples)</li>
</div>
</div>
<div class="row">
<div class="col-md-12">
<li><a
href="https://awma1-my.sharepoint.com/:u:/g/personal/yuz_l0_tn/EdIUQBXaNXVMm8-Kmvg3gjcBHmnaeE5s4OP8OO5Ics7URA?download=1"
class="download-link">Test questions</a> (1,503,510 samples)</li>
</div>
</div>
<div class="row">
<div class="col-md-12">
<li><a
href="https://awma1-my.sharepoint.com/:u:/g/personal/yuz_l0_tn/ERBudANdq9JAgXHBw_YsnZsBso6nskwG9FxyShhxhSQ3Tg?download=1"
class="download-link">Test-dev questions</a> (300,694 samples)</li>
</div>
</div>
<div class="row">
<div class="col-md-12">
<li><a
href="https://awma1-my.sharepoint.com/:u:/g/personal/yuz_l0_tn/EYiwgEwCXu5Apcoj3I-Y3ewBBHPb0N24po5q13Ep9BXPFA?download=1"
class="download-link">Test-tiny questions</a> (20,000 samples)</li>
*The <i>test-dev</i> and <i>test-tiny</i> splits are two subsets of the <i>test</i> split.
</div>
</div>
<div class="row">
<div class="col-md-9" style="margin-top:20px; width: 500px; ">
More details of the dataset are provided <a
href="https://github.com/MILVLG/anetqa-code/tree/main/dataset">here</a>.
</div>
</div>
</div>
</div>
<br><br>
<div class="container">
<h2 id="Code" style="padding-top: 80px; margin-top: -80px;">Code</h2>
<hr>
<div class="row">
<div class="col-md-9">
Code for ANetQA baseline models are available <a href="https://github.com/MILVLG/anetqa-code">here</a>.
</div>
</div>
</div><br><br>
<div class="container">
<h2 id="Eval" style="padding-top: 80px; margin-top: -80px;">Evaluation</h2>
<hr>
<div class="row">
<div class="col-md-9">
Evaluation for the testing set is provided on the online <a href="https://eval.ai/web/challenges/challenge-page/2226/overview">EvalAI server</a>.
</div>
</div>
<h3>Submit Format</h3>
<p style="font-size:15px; font-weight: 200; border-style: solid;
border-width: 1px; text-align:justify;">
<code style="background-color: #fff;">
[...<br>
{<br>
"question_id": question_id,<br>
"answer": answer<br>
},<br>
...]
</code>
</p>
We have provided an example result JSON file <a
href="https://github.com/MILVLG/anetqa-code/blob/main/dataset/fake_res.json">here</a>.
</div><br><br>
<div class="container">
<h2 id="Licence" style="padding-top: 80px; margin-top: -80px;">Licence</h2>
<hr>
<div class="row">
<div class="col-md-9">
The annotations in this dataset belong to the ANetQA Team and are licensed under a <a href="https://creativecommons.org/licenses/by-nc/4.0/deed.en">CC BY-NC 4.0 </a>License.
</div>
</div>
</div><br><br>
<div class="container">
<h2 id="More" style="padding-top: 80px; margin-top: -80px;">Bibtex</h2>
<hr>
<div class="row">
<pre style="font-size:12px;margin-left: 15px;margin-right: 15px;">
@inproceedings{yu2023anetqa,
title={ANetQA: A Large-scale Benchmark for Fine-grained Compositional Reasoning over Untrimmed Videos},
author={Yu, Zhou and Zheng, Lixiang and Zhao, Zhou and Wu, Fei and Fan, Jianping and Ren, Kui and Yu, Jun},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
pages={23191--23200}
year={2023}
}</pre>
</div>
</div><br><br>
</div>
</main>
</body>
<div style="all: initial;">
<div id="__hcfy__" style="all: initial;"></div>
</div>
</html>