-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathflowfeat.html
More file actions
348 lines (295 loc) · 15.8 KB
/
flowfeat.html
File metadata and controls
348 lines (295 loc) · 15.8 KB
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
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0, shrink-to-fit=no">
<title>FlowFeat: Pixel-Dense Embedding of Motion Profiles</title>
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/twitter-bootstrap/4.5.0/css/bootstrap.min.css">
<link href='https://fonts.googleapis.com/css?family=Source+Sans+Pro:300,400,500,600' rel='stylesheet' type='text/css'>
<link rel="stylesheet" href="/assets/css/Highlight-Clean.css">
<link rel="stylesheet" href="/assets/css/styles.css">
<link rel="apple-touch-icon" sizes="180x180" href="/assets/images/flowfeat/apple-touch-icon.png">
<link rel="icon" type="image/png" sizes="32x32" href="/assets/images/flowfeat/favicon-32x32.png">
<link rel="icon" type="image/png" sizes="16x16" href="/assets/images/flowfeat/favicon-16x16.png">
<meta property="og:site_name" content="FlowFeat: Pixel-Dense Embedding of Motion Profiles" />
<meta property="og:type" content="video.other" />
<meta property="og:title" content="FlowFeat: Pixel-Dense Embedding of Motion Profiles" />
<meta property="og:description" content="Distilling optical flow networks into pixel-level task-agnostic representations." />
<meta property="og:url" content="https://tum-vision.github.io/flowfeat" />
<meta property="og:image" content="https://tum-vision.github.io/flowfeat/assets/images/flowfeat_samples.png" />
<meta property="article:publisher" content="https://tum-vision.github.io/flowfeat" />
<meta name="twitter:card" content="summary_large_image" />
<meta name="twitter:title" content="FlowFeat: Pixel-Dense Embedding of Motion Profiles" />
<meta name="twitter:description" content="Distilling optical flow networks into pixel-level task-agnostic representations." />
<meta name="twitter:url" content="https://tum-vision.github.io/flowfeat" />
<meta name="twitter:image" content="https://tum-vision.github.io/flowfeat/assets/images/flowfeat_samples.png" />
<!-- <meta name="twitter:site" content="" /> -->
<script type="module" src="https://unpkg.com/@google/model-viewer@2.0.1/dist/model-viewer.min.js"></script>
</head>
<body>
<div class="highlight-clean" style="padding-bottom: 10px;">
<div class="container" style="max-width: 768px;">
<h1 class="text-center"><b>FlowFeat</b>:<br> Pixel-Dense Embedding of Motion Profiles</h1>
</div>
<br/>
<div class="container" style="max-width: 768px;">
<div class="row authors">
<div class="col-sm-4">
<h5 class="text-center"><a class="text-center" href="https://arnike.github.io">Nikita Araslanov</a></h5>
<h6 class="text-center">TU Munich / MCML</h6>
</div>
<div class="col-sm-4">
<h5 class="text-center"><a href="#">Anna Sonnweber</a></h5>
<h6 class="text-center">TU Munich</h6>
</div>
<div class="col-sm-4">
<h5 class="text-center"><a class="text-center" href="https://cvg.cit.tum.de/members/cremers">Daniel Cremers</a></h5>
<h6 class="text-center">TU Munich / MCML</h6>
</div>
</div>
</div>
<br/>
<h1 class="text-center" style="font-size:22px;">✨ NeurIPS 2025 Spotlight ✨</h1>
<br/>
<div class="buttons" style="margin-bottom: 8px;">
<a class="btn btn-light" role="button" href="https://cvg.cit.tum.de/_media/research/flowfeat/camera_ready.pdf">
<svg style="width:24px;height:24px;margin-left:-12px;margin-right:12px" viewBox="0 0 24 24">
<path fill="currentColor" d="M16 0H8C6.9 0 6 .9 6 2V18C6 19.1 6.9 20 8 20H20C21.1 20 22 19.1 22 18V6L16 0M20 18H8V2H15V7H20V18M4 4V22H20V24H4C2.9 24 2 23.1 2 22V4H4M10 10V12H18V10H10M10 14V16H15V14H10Z"></path>
</svg>Paper
</a>
<a class="btn btn-light" role="button" href="https://cvg.cit.tum.de/_media/research/flowfeat/camera_ready_supp.zip">
<svg style="width:24px;height:24px;margin-left:-12px;margin-right:12px" viewBox="0 0 24 24">
<path fill="currentColor" stroke-width="1" d="M14 2H6a2 2 0 0 0-2 2v16c0 1.1.9 2 2 2h12c1.1 0 2-.9 2-2V8l-6-6Zm1 7V3.5L19.5 9H15ZM10 8h2v2h-2V8Zm0 3h2v2h-2v-2Zm0 3h2v2h-2v-2Zm1 3c.55 0 1 .45 1 1v1h-2v-1c0-.55.45-1 1-1Z"></path>
</svg>Supplemental Material
</a>
<a class="btn btn-light" aria-disabled="true" role="button" href="https://github.com/tum-vision/flowfeat">
<svg style="margin-left:-12px;margin-right:12px" xmlns="http://www.w3.org/2000/svg" width="24" height="24" fill="currentColor" class="bi bi-github" viewBox="0 0 16 16">
<path d="M8 0C3.58 0 0 3.58 0 8c0 3.54 2.29 6.53 5.47 7.59.4.07.55-.17.55-.38 0-.19-.01-.82-.01-1.49-2.01.37-2.53-.49-2.69-.94-.09-.23-.48-.94-.82-1.13-.28-.15-.68-.52-.01-.53.63-.01 1.08.58 1.23.82.72 1.21 1.87.87 2.33.66.07-.52.28-.87.51-1.07-1.78-.2-3.64-.89-3.64-3.95 0-.87.31-1.59.82-2.15-.08-.2-.36-1.02.08-2.12 0 0 .67-.21 2.2.82.64-.18 1.32-.27 2-.27s1.36.09 2 .27c1.53-1.04 2.2-.82 2.2-.82.44 1.1.16 1.92.08 2.12.51.56.82 1.27.82 2.15 0 3.07-1.87 3.75-3.65 3.95.29.25.54.73.54 1.48 0 1.07-.01 1.93-.01 2.2 0 .21.15.46.55.38A8.01 8.01 0 0 0 16 8c0-4.42-3.58-8-8-8"/>
</svg>
Code
</a>
</div>
</div>
<hr class="divider" />
<div class="container" style="max-width: 768px;">
<div class="row">
<div class="col-md-12" style="font-size: 16pt;">
<b>The gist:</b>
<span>
<!-- <strong> -->
Distilling optical flow networks into pixel-level task-agnostic representations.
<!-- </strong> -->
</span>
</div>
</div>
</div>
<br/>
<div class="container text-center my-4">
<!-- Row 1 captions -->
<div class="row small text-muted mb-2">
<div class="col"><strong>Input</strong></div>
<div class="col">DINOv2</div>
<div class="col">V-JEPA</div>
<div class="col">VideoMAE</div>
<div class="col">FeatUp</div>
<div class="col"><strong>FlowFeat</strong></div>
</div>
<!-- Row 1 images -->
<div class="row mb-2">
<div class="col p-1"><img src="assets/images/flowfeat/panda_input.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/dinov2/panda.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/vjepa/panda-vjepa.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/videomae/panda-vmae.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/featup/panda.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/panda_pca.png" class="img-fluid"></div>
</div>
<!-- Row 2 images -->
<div class="row g-0 mb-2">
<div class="col p-1"><img src="assets/images/flowfeat/turtle_input.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/dinov2/turtle.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/vjepa/turtle-vjepa.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/videomae/turtle-vmae.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/featup/turtle.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/turtle_pca.png" class="img-fluid"></div>
</div>
<!-- Row 3 images -->
<div class="row g-0 mb-2">
<div class="col p-1"><img src="assets/images/flowfeat/skateboard_input.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/dinov2/skateboard.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/vjepa/skateboard-vjepa.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/videomae/skateboard-vmae.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/featup/skateboard.png" class="img-fluid"></div>
<div class="col p-1"><img src="assets/images/flowfeat/skateboard_pca.png" class="img-fluid"></div>
</div>
<!-- Figure caption -->
<h6 class="caption">
<strong>FlowFeat</strong> provides versatile pixel-level features. Using motion-driven embedding statistics, it achieves high spatial precision and temporal consistency.
Shown: PCA visualization of feature maps vs. state-of-the-art encoders.
</h6>
</div>
<hr class="divider" />
<h3 class="text-center mb-3">More FlowFeat Examples</h3>
<div class="container mt-4">
<div class="row g-0 mb-2">
<div class="col p-1">
<div class="hover-img">
<img src="assets/images/flowfeat/examples/snakes_input.png" class="main">
<img src="assets/images/flowfeat/examples/snakes_pca.png" class="hover">
<div class="img-title">
<span>Hover to view</span>
</div>
</div>
</div>
<div class="col p-1">
<div class="hover-img">
<img src="assets/images/flowfeat/examples/piano_input.png" class="main">
<img src="assets/images/flowfeat/examples/piano_pca.png" class="hover">
<div class="img-title">
<span>Hover to view</span>
</div>
</div>
</div>
<div class="col p-1">
<div class="hover-img">
<img src="assets/images/flowfeat/examples/deer_input.png" class="main">
<img src="assets/images/flowfeat/examples/deer_pca.png" class="hover">
<div class="img-title">
<span>Hover to view</span>
</div>
</div>
</div>
<div class="col p-1">
<div class="hover-img">
<img src="assets/images/flowfeat/examples/turtles_input.png" class="main">
<img src="assets/images/flowfeat/examples/turtles_pca.png" class="hover">
<div class="img-title">
<span>Hover to view</span>
</div>
</div>
</div>
<div class="col p-1">
<div class="hover-img">
<img src="assets/images/flowfeat/examples/skater_input.png" class="main">
<img src="assets/images/flowfeat/examples/skater_pca.png" class="hover">
<div class="img-title">
<span>Hover to view</span>
</div>
</div>
</div>
<div class="col p-1">
<div class="hover-img">
<img src="assets/images/flowfeat/examples/zebras_input.png" class="main">
<img src="assets/images/flowfeat/examples/zebras_pca.png" class="hover">
<div class="img-title">
<span>Hover to view</span>
</div>
</div>
</div>
</div>
</div>
<hr class="divider" />
<h2 class="text-center mb-3">Video Object Segmentation</h2>
<div class="container text-center" style="max-width: 768px;">
<div class="row">
<div class="col-md-12">
<video class="video lazy" autoplay loop playsinline controls muted>
<source src="https://cvg.cit.tum.de/_media/research/flowfeat/a_vos_featup.mp4" type="video/mp4"></source>
</video>
<h6 class="caption">
<strong>FlowFeat</strong> is a fine-grained feature representation, where dynamic objects are particularly prominent.
</h6>
</div>
</div>
<br/>
<div class="row">
<div class="col-md-12">
<h3>Comparison to FeatUp (Fu et al., 2024)</h3>
<video class="video lazy" autoplay loop playsinline controls muted>
<source src="https://cvg.cit.tum.de/_media/research/flowfeat/b_pca_featup.mp4" type="video/mp4"></source>
</video>
<h6 class="caption">
<strong>FlowFeat</strong> scales well with the input resolution – here, increased by a factor of 2.
</h6>
</div>
</div>
<br/>
<div class="row">
<div class="col-md-12">
<h3>Comparison to LoftUp (Huang et al., 2025)</h3>
<video class="video lazy" autoplay loop playsinline controls muted>
<source src="https://cvg.cit.tum.de/_media/research/flowfeat/d_pca_vs_loftup.mp4" type="video/mp4"></source>
</video>
<h6 class="caption">
FlowFeat fares well against supervised models such as LoftUp and can be trained in a completely unsupervised fashion.
</h6>
</div>
</div>
</div>
<hr class="divider" />
<h2 class="text-center mb-3">Citation</h2>
<div class="container" style="max-width: 768px;">
<div class="row">
<div class="col-md-12">
<code>
@inproceedings{Araslanov:2025:FlowFeat,<br>
author = {Araslanov, Nikita and Sonnweber, Anna and Cremers, Daniel},<br>
title = {{FlowFeat}: Pixel-Dense Embedding of Motion Profiles},<br>
booktitle = {NeurIPS},<br>
year = {2025},<br>
}</code>
</div>
</div>
</div>
<hr class="divider" />
<div class="container" style="max-width: 768px;">
<footer>
Website template credit: <a href="https://dreamfusion3d.github.io/">DreamFusion</a>
<footer>
</div>
<!-- Cookie Consent Banner -->
<div id="cookie-banner"
class="alert alert-dark alert-dismissible fade fixed-bottom mb-0"
role="alert"
style="display:none;">
<div class="container d-flex flex-column align-items-center text-center">
<span class="mb-2">This site uses cookies for analytics.</span>
<button id="accept-cookies" class="btn btn-primary btn-sm">Accept</button>
</div>
</div>
<script src="https://cdnjs.cloudflare.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/twitter-bootstrap/4.5.0/js/bootstrap.bundle.min.js"></script>
<!-- GA -->
<script>
/* Load Google Analytics AFTER consent */
function loadGA() {
var s = document.createElement('script');
s.async = true;
s.src = "https://www.googletagmanager.com/gtag/js?id=G-1HCLZM84F4";
document.head.appendChild(s);
window.dataLayer = window.dataLayer || [];
function gtag(){ dataLayer.push(arguments); }
window.gtag = gtag;
gtag('js', new Date());
gtag('config', 'G-1HCLZM84F4');
}
/* On document ready */
$(function () {
if (!localStorage.getItem('cookiesAccepted')) {
$('#cookie-banner').addClass('show').show();
} else {
loadGA();
}
$('#accept-cookies').on('click', function () {
localStorage.setItem('cookiesAccepted', 'true');
loadGA();
$('#cookie-banner')
.removeClass('show')
.alert('close'); // Bootstrap dismiss
});
$('.hover-img').one('mouseenter touchstart', function() {
$(this).addClass('hide-hint');
});
});
</script>
</body>
</html>