forked from facebookresearch/Motionformer
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
376 lines (337 loc) · 13.9 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
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<!-- ======================================================================= -->
<script src="http://www.google.com/jsapi" type="text/javascript"></script>
<script type="text/javascript">google.load("jquery", "1.3.2");</script>
<link rel="stylesheet" href="https://maxcdn.bootstrapcdn.com/bootstrap/4.5.0/css/bootstrap.min.css">
<script src="https://maxcdn.bootstrapcdn.com/bootstrap/4.5.0/js/bootstrap.min.js"></script>
<style type="text/css">
body {
/* font-family: "Titillium Web","HelveticaNeue-Light", "Helvetica Neue Light", "Helvetica Neue", Helvetica, Arial, "Lucida Grande", sans-serif; */
font-family: "Avenir", sans-serif;
font-weight:400;
font-size:18px;
margin-left: auto;
margin-right: auto;
width: 1100px;
}
h1 {
font-weight:300;
}
.disclaimerbox {
background-color: #eee;
border: 1px solid #eeeeee;
border-radius: 10px ;
-moz-border-radius: 10px ;
-webkit-border-radius: 10px ;
padding: 20px;
}
video.header-vid {
height: 140px;
border: 1px solid black;
border-radius: 10px ;
-moz-border-radius: 10px ;
-webkit-border-radius: 10px ;
}
img.header-img {
height: 140px;
border: 1px solid black;
border-radius: 10px ;
-moz-border-radius: 10px ;
-webkit-border-radius: 10px ;
}
img.rounded {
border: 1px solid #eeeeee;
border-radius: 10px ;
-moz-border-radius: 10px ;
-webkit-border-radius: 10px ;
}
a:link,a:visited
{
color: #1367a7;
text-decoration: none;
}
a:hover {
color: #208799;
}
td.dl-link {
height: 160px;
text-align: center;
font-size: 22px;
}
.layered-paper-big { /* modified from: http://css-tricks.com/snippets/css/layered-paper/ */
box-shadow:
0px 0px 1px 1px rgba(0,0,0,0.35), /* The top layer shadow */
5px 5px 0 0px #fff, /* The second layer */
5px 5px 1px 1px rgba(0,0,0,0.35), /* The second layer shadow */
10px 10px 0 0px #fff, /* The third layer */
10px 10px 1px 1px rgba(0,0,0,0.35), /* The third layer shadow */
15px 15px 0 0px #fff, /* The fourth layer */
15px 15px 1px 1px rgba(0,0,0,0.35), /* The fourth layer shadow */
20px 20px 0 0px #fff, /* The fifth layer */
20px 20px 1px 1px rgba(0,0,0,0.35), /* The fifth layer shadow */
25px 25px 0 0px #fff, /* The fifth layer */
25px 25px 1px 1px rgba(0,0,0,0.35); /* The fifth layer shadow */
margin-left: 10px;
margin-right: 45px;
}
.layered-paper { /* modified from: http://css-tricks.com/snippets/css/layered-paper/ */
box-shadow:
0px 0px 1px 1px rgba(0,0,0,0.35), /* The top layer shadow */
5px 5px 0 0px #fff, /* The second layer */
5px 5px 1px 1px rgba(0,0,0,0.35), /* The second layer shadow */
10px 10px 0 0px #fff, /* The third layer */
10px 10px 1px 1px rgba(0,0,0,0.35); /* The third layer shadow */
margin-top: 5px;
margin-left: 10px;
margin-right: 30px;
margin-bottom: 5px;
}
.vert-cent {
position: relative;
top: 50%;
transform: translateY(-50%);
}
hr
{
border: 0;
height: 1px;
background-image: linear-gradient(to right, rgba(0, 0, 0, 0), rgba(0, 0, 0, 0.75), rgba(0, 0, 0, 0));
}
#authors td {
padding-bottom:5px;
padding-top:30px;
}
</style>
<html>
<head>
<title>Motionformer</title>
<meta property="og:title" content="Keeping Your Eye on the Ball: Trajectory Attention in Video Transformers " />
<meta property="og:image" content="" />
<meta property="og:image:width" content="3000" />
<meta property="og:image:height" content="800" />
</head>
<body>
<br><br><br><br>
<center>
<span style="font-size:38px">Keeping Your Eye on the Ball: Trajectory Attention in Video Transformers</span>
<br><br>
<table align=center width=800px>
<tr>
<td align=center width=200px>
<center>
<span style="font-size:24px"><a href="https://scholar.google.com/citations?hl=en&user=KOURpBsAAAAJ">Mandela Patrick</a></span>
</center>
</td>
<td align=center width=200px>
<center>
<span style="font-size:24px"><a href="https://sites.google.com/view/djcampbell">Dylan Campbell</a></span>
</center>
</td>
<td align=center width=200px>
<center>
<span style="font-size:24px"><a href="https://yukimasano.github.io/">Yuki Asano</a></span>
</center>
</td>
<tr>
<td align=center width=200px>
<center>
<span style="font-size:24px"><a href="https://imisra.github.io/">Ishan Misra</a></span>
</center>
</td>
<td align=center width=200px>
<center>
<span style="font-size:24px"><a href="https://www.cs.cmu.edu/~fmetze/interACT/Home.html">Florian Metze</a></span>
</center>
</td>
<td align=center width=200px>
<center>
<span style="font-size:24px"><a href="https://feichtenhofer.github.io/">Christoph Feichtenhofer</a></span>
</center>
</td>
<tr>
<td align=center width=300px>
<center>
<span style="font-size:24px"><a href="https://www.robots.ox.ac.uk/~vedaldi/">Andrea Vedaldi</a></span>
</center>
</td>
<td align=center width=300px>
<center>
<span style="font-size:24px"><a href="https://www.robots.ox.ac.uk/~joao/">Joao Henriques</a></span>
</center>
</td>
</table>
<br><br>
<table align=center width=650px style="font-size:20px;">
<tr>
<td align=center width=150px>
<center>
<span><a href="https://arxiv.org/abs/2106.05392"> [Paper]</a></span>
</center>
</td>
<td align=center width=150px>
<center>
<span style=""><a href="https://github.com/facebookresearch/Motionformer"> [Code]</a></span>
</center>
</td>
</tr>
<tr>
</table>
<br>
</center>
<br><br>
<table align=center width=850px>
<tr>
<td width=400px>
<center>
<a href="figs/spash.png"><img src = "figs/splash.png" width="90%"></img></href></a><br>
</center>
</td>
</tr>
</table>
<br><br>
<hr>
<table align=center>
<div class='row'>
<div class='column'>
<center>
<h1>Abstract</h1>
<p style="width:560px; text-align:left; color:darkslategray; ;">In video transformers, the time dimension is often treated in the same way as the two spatial dimensions.
However, in a scene where objects or the camera may move, a physical point imaged at one location in frame t may be entirely unrelated to what is found at that location in frame t+k.
These temporal correspondences should be modeled to facilitate learning about dynamic scenes.
To this end, we propose a new drop-in block for video transformers---trajectory attention---that aggregates information along implicitly determined motion paths.
We additionally propose a new method to address the quadratic dependence of computation and memory on the input size, which is particularly important for high resolution or long videos.
While these ideas are useful in a range of settings, we apply them to the specific task of video action recognition with a transformer model and obtain state-of-the-art results on the Kinetics, Something--Something V2, and Epic-Kitchens datasets. </p>
</center>
</div>
<div class="column" style='float:right; width:40%'>
<center>
<h1>Trajectory Attention</h1>
<!-- <embed src="figs/pseudocode.pdf" width="800px" height="2100px" /> -->
<img src = "figs/traj_attn_fig.png" width="90%"></img>
</center>
</div>
</div>
</table>
<br><br>
</table>
<hr>
<center>
<h1>Quantitative Results</h1>
<table class="table table-responsive">
<caption style="text-align:center; caption-side:top"> <b> Something-Something V2</caption>
<thead>
<tr>
<th align=center width=350px>Model</th>
<th align=center width=350px>Top-1</th>
<th align=center width=350px>Top-5</th>
</tr>
</thead>
<tbody>
<tr>
<td style="text-align:center" width=350px>TimeSformer-HR</td>
<td style="text-align:center" width=350px>62.5</td>
<td style="text-align:center" width=350px>-</td>
</tr>
<tr>
<td style="text-align:center" width=350px>ViViT-L</td>
<td style="text-align:center" width=350px>65.4</td>
<td style="text-align:center" width=350px>89.8</td>
</tr>
<tr>
<td style="text-align:center" width=350px>MViT-B</td>
<td style="text-align:center" width=350px>67.1</td>
<td style="text-align:center" width=350px>90.8</td>
</tr>
<tr>
<td style="text-align:center" width=350px> <b> Motionformer-L</td>
<td style="text-align:center" width=350px> <b> 68.1</td>
<td style="text-align:center" width=350px> <b> 91.2</td>
</tr>
</tbody>
</table>
<table class="table table-responsive">
<caption style="text-align:center; caption-side:top"> <b> Kinetics-400</caption>
<thead>
<tr>
<th align=center width=350px>Model</th>
<th align=center width=350px>Top-1</th>
<th align=center width=350px>Top-5</th>
</tr>
</thead>
<tbody>
<tr>
<td style="text-align:center" width=350px>TimeSformer-HR</td>
<td style="text-align:center" width=350px>80.7</td>
<td style="text-align:center" width=350px>94.7</td>
</tr>
<tr>
<td style="text-align:center" width=350px>MViT-B</td>
<td style="text-align:center" width=350px>81.2</td>
<td style="text-align:center" width=350px>95.1</td>
</tr>
<tr>
<td style="text-align:center" width=350px>ViViT-L</td>
<td style="text-align:center" width=350px> <b> 81.3</td>
<td style="text-align:center" width=350px>94.7</td>
</tr>
<tr>
<td style="text-align:center" width=350px> <b> Motionformer-HR</td>
<td style="text-align:center" width=350px> 81.1</td>
<td style="text-align:center" width=350px> <b> 95.2</td>
</tr>
</tbody>
</table>
</center>
<hr>
<center>
<h1>Qualitative Results</h1>
<table border="0" align="center" cellspacing="0" cellpadding="20">
<td align="center" valign="middle">
<img src = "figs/qual_results.png" width="90%"></img>
</td>
</table>
</center>
<hr>
<!-- <table align=center width=550px> -->
<table align=center width=800px>
<center><h1>Paper</h1></center>
<tr>
<td><a href=""><img class="layered-paper-big" style="height:175px" src="./figs/firstpage.png"/></a></td>
<td><span style="font-size:14pt">Mandela Patrick, Dylan Campbell, Yuki Asano, Ishan Misra, Florian Metze, Christoph Feichtenhofer, Andrea Vedaldi, Joao Henriques<br>
<i><a href="https://arxiv.org/abs/2106.05392">Keeping Your Eye on the Ball: Trajectory Attention in Video Transformers</a></i><br>
ArXiv<br>
</a>
<span style="font-size:4pt"><a href=""><br></a>
</span>
</td>
</td>
</tr>
</table>
<br><br>
<hr>
<table align=center width=1100px>
<tr>
<td width=400px>
<left>
<center><h1>Acknowledgements</h1></center>
We thank Bernie Huang, Dong Guo, Rose Kanjirathinkal, Gedas Bertasius, Mike Zheng Shou, Mathilde Caron, Hugo Touvron, Benjamin Lefaudeux, Haoqi Fan, and Geoffrey Zweig from Facebook AI for their help, support, and discussion around this project.
We also thank Max Bain and Tengda Han from VGG for fruitful discussions.
<br><br><br><br>
<center>
<footer id="footer">
<p class="copyright"><font size="-1"> Website design adapted
from <a href="https://ajabri.github.io/videowalk/">Allan
Jabri </a></font></p><font size="-1"><a href="https://ajabri.github.io/videowalk/">
</a></font></footer><font size="-1"><a href="https://ajabri.github.io/videowalk/">
</a></font></center>
</table>
<br><br>
</table>
</left>
</td>
</tr>
</table>
<br><br>
</body>
</html>