-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
626 lines (543 loc) · 31.1 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
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
<!DOCTYPE html>
<html>
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Video-RAG</title>
<link rel="icon" href="./static/images/icon.png">
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" rel="stylesheet">
<link rel="stylesheet" href="./static/css/bulma.min.css">
<link rel="stylesheet" href="./static/css/bulma-carousel.min.css">
<link rel="stylesheet" href="./static/css/bulma-slider.min.css">
<link rel="stylesheet" href="./static/css/fontawesome.all.min.css">
<link rel="stylesheet" href="./static/css/index.css">
<link rel="stylesheet" href="./static/css/leaderboard.css">
<link rel="stylesheet" href="./static/css/video-player.css">
<script type="text/javascript" src="static/js/sort-table.js" defer></script>
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script defer src="./static/js/fontawesome.all.min.js"></script>
<script src="./static/js/bulma-carousel.min.js"></script>
<script src="./static/js/bulma-slider.min.js"></script>
<script src="./static/js/explorer-index.js"></script>
<script src="./static/js/question_card.js"></script>
<script src="./static/js/leaderboard_testmini.js"></script>
<script src="./data/results/output_folders.js" defer></script>
<script src="./data/results/model_scores.js" defer></script>
<script src="./visualizer/data/data_public.js" defer></script>
<style>
.no-sort {
cursor: default;
pointer-events: none;
background-image: none !important; /* Remove the sort arrow */
}
</style>
</head>
<body>
<nav class="navbar" role="navigation" aria-label="main navigation">
<div class="navbar-brand">
<a role="button" class="navbar-burger" aria-label="menu" aria-expanded="false">
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
</a>
</div>
<div class="navbar-menu">
<div class="navbar-start" style="flex-grow: 1; justify-content: center;">
<div class="navbar-item has-dropdown is-hoverable">
<p style="font-size:18px; display: inline; margin-right: -2px; margin-top: 12px;">🔥</p>
<a class="navbar-link">
More Research
</a>
<div class="navbar-dropdown">
<a class="navbar-item" href="https://github.com/BradyFU/Video-MME">
<b>Video-MME</b>
</a>
<a class="navbar-item" href="https://github.com/Leon1207/3DRefTR">
<b>3DRefTR</b>
</a>
<a class="navbar-item" href="https://github.com/Leon1207/3DGCTR">
<b>3DGCTR</b>
</a>
</div>
</div>
</div>
</div>
</nav>
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<h1 class="title is-1 publication-title is-bold">
<span class="video-mme" style="vertical-align: middle">Video-RAG</span>
</h1>
<h2 class="subtitle is-3 publication-subtitle", style="margin-bottom: 20px;">
Visually-aligned Retrieval-Augmented <br> Long Video Comprehension
</h2>
<div class="is-size-5 publication-authors">
<span class="author-block">Yongdong Luo<sup style="color:#6fbf73;">1</sup></span>
<span class="author-block">Xiawu Zheng<sup style="color:#6fbf73;">1</sup></span>
<span class="author-block">Xiao Yang<sup style="color:#6fbf73;">1</sup></span>
<span class="author-block">Guilin Li<sup style="color:#6fbf73;">1</sup></span>
<span class="author-block">Haojia Lin<sup style="color:#6fbf73;">1</sup></span><br>
<span class="author-block">Jinfa Huang<sup style="color:#ed4b82">2</sup></span>
<span class="author-block">Jiayi Ji<sup style="color:#6fbf73;">1</sup></span>
<span class="author-block">Fei Chao<sup style="color:#6fbf73;">1</sup></span>
<span class="author-block">Jiebo Luo<sup style="color:#ed4b82">2</sup></span>
<span class="author-block">Rongrong Ji<sup style="color:#6fbf73;">1</sup></span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block"><sup style="color:#6fbf73;">1</sup>Xiamen University,</span>
<span class="author-block"><sup style="color:#ed4b82">2</sup>University of Rochester</span>
<!-- <span class="paper-block"><b style="color:#f41c1c">ICLR 2024 Oral</b> (85 in 7304, 1.2%)</span> -->
</div>
<div class="column has-text-centered">
<div class="publication-links">
<!-- PDF Link. -->
<span class="link-block">
<a href="https://arxiv.org/abs/2411.13093"
class="external-link button is-normal is-rounded is-dark">
<!-- <a href="https://lupantech.github.io/papers/arxiv23_mathvista.pdf"
class="external-link button is-normal is-rounded is-dark"> -->
<span class="icon">
<i class="fas fa-file-alt"></i>
</span>
<span>arXiv</span>
</a>
</span>
<!-- Code Link. -->
<span class="link-block">
<a href="https://github.com/Leon1207/Video-RAG-master"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container" style="margin-bottom: 2vh; margin-top: -70px;">
<!-- demo -->
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Video Demo</h2>
<div class="content has-text-justified">
<video width="1920" height="780" controls autoplay loop>
<source src="./static/images/demo.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
<p>
Introduction from the podcast AI Agents Unleashed: https://www.youtube.com/watch?v=WTs3xHicR_0
</p>
</div>
</div>
</div>
<!-- Abstract. -->
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Introduction</h2>
<div class="content has-text-justified">
Existing large video-language models (LVLMs) struggle to comprehend long videos correctly due to limited context.
To address this problem, fine-tuning long-context LVLMs and employing GPT-based agents have emerged as promising solutions.
However, fine-tuning LVLMs would require extensive high-quality data and substantial GPU resources,
while GPT-based agents would rely on proprietary models (e.g., GPT-4o). In this paper,
we propose Video Retrieval-Augmented Generation (Video-RAG),
a training-free and cost-effective pipeline that employs visually-aligned auxiliary texts to help facilitate cross-modality alignment while providing additional information beyond the visual content.
Specifically, we leverage open-source external tools to extract visually-aligned information from pure video data (e.g., audio, optical character, and object detection), and incorporate the extracted information into an existing LVLM as auxiliary texts, alongside video frames and queries, in a plug-and-play manner.
Our Video-RAG offers several key advantages:
(i) lightweight with low computing overhead due to single-turn retrieval;
(ii) easy implementation and compatibility with any LVLM;
and (iii) significant, consistent performance gains across long video understanding benchmarks, including Video-MME, MLVU, and LongVideoBench. Notably, our model demonstrates superior performance over proprietary models like Gemini-1.5-Pro and GPT-4o when utilized with a 72B model.
</div>
</div>
</div>
<!--/ Abstract. -->
</div>
</section>
<script>
document.addEventListener("DOMContentLoaded", function() {
updateRowNumbers();
document.querySelector("#results").addEventListener("click", updateRowNumbers);
});
function updateRowNumbers() {
const rows = document.querySelectorAll("#results tbody tr");
rows.forEach((row, index) => {
row.querySelector("td").innerText = index + 1;
});
}
</script>
<section class="section">
<div class="container">
<div class="columns is-centered">
<div class="column is-full has-text-centered content">
<h2 class="title is-3" id="leaderboard">Experimental results on Video-MME</h2>
<div class="content">
<p>
By default, this leaderboard is sorted by results with Video-RAG. To view other sorted results, please click on the corresponding cell.
</p>
<p>
The results are kept on updating!
</p>
</div>
<table class="js-sort-table" id="results" style="margin-left: auto; margin-right: auto;">
<thead>
<tr>
<th rowspan="2" class="no-sort" style="vertical-align: middle; width: 50px;" data-js-sort-colNum="none"><strong>#</strong></th>
<th rowspan="2" class="no-sort" style="vertical-align: middle; width: 180px;" data-js-sort-colNum="none"><strong>Model</strong></th>
<th rowspan="2" class="no-sort" style="vertical-align: middle; width: 50px;" data-js-sort-colNum="none"><strong>LLM <br>Params</strong></th>
<th rowspan="2" class="no-sort" style="vertical-align: middle; width: 100px;" data-js-sort-colNum="none"><strong>Frames</strong></th>
<th colspan="3" style="vertical-align: middle; width: 300px;" data-js-sort-colNum="6"><strong>Overall (%)</strong></th>
<th colspan="3" style="vertical-align: middle; width: 300px;" data-js-sort-colNum="9"><strong>Short Video (%)</strong></th>
<th colspan="3" style="vertical-align: middle; width: 300px;" data-js-sort-colNum="12"><strong>Medium Video (%)</strong></th>
<th colspan="3" style="vertical-align: middle; width: 300px;" data-js-sort-colNum="15"><strong>Long Video (%)</strong></th>
</tr>
<tr>
<th data-js-sort-colNum="4"><strong>w/o subs</strong></th>
<th data-js-sort-colNum="5"><strong>w sub</strong></th>
<th data-js-sort-colNum="6"><strong>w Ours</strong></th>
<th data-js-sort-colNum="7"><strong>w/o sub</strong></th>
<th data-js-sort-colNum="8"><strong>w subs</strong></th>
<th data-js-sort-colNum="9"><strong>w Ours</strong></th>
<th data-js-sort-colNum="10"><strong>w/o sub</strong></th>
<th data-js-sort-colNum="11"><strong>w subs</strong></th>
<th data-js-sort-colNum="12"><strong>w Ours</strong></th>
<th data-js-sort-colNum="13"><strong>w/o sub</strong></th>
<th data-js-sort-colNum="14"><strong>w subs</strong></th>
<th data-js-sort-colNum="15"><strong>w Ours</strong></th>
</tr>
</thead>
<tbody>
<tr>
<td style="vertical-align: middle;"></td>
<td style="text-align: left; padding: 2px 10px;">
<b class="">
<a href="https://storage.googleapis.com/deepmind-media/gemini/gemini_v1_5_report.pdf" class="ext-link" style="font-size: 16px; margin-left: 5px;">Gemini 1.5 Pro</a>
<p style="font-size: 12px; margin-left: 5px; color: #858383;">Google</p>
</b>
</td>
<td style="vertical-align: middle;">- </td>
<td style="vertical-align: middle;">1/0.5 fps </td>
<td style="vertical-align: middle;"><b class="">75.0</b></td>
<td style="vertical-align: middle;"><b class="">81.3</b></td>
<td style="vertical-align: middle;"><strong style="color: #22a56e;" >-</strong></td>
<td style="vertical-align: middle;">81.7</td>
<td style="vertical-align: middle;">84.5</td>
<td style="vertical-align: middle;"><text style="color: #22a56e;" >- </text></td>
<td style="vertical-align: middle;">74.3</td>
<td style="vertical-align: middle;">81.0</td>
<td style="vertical-align: middle;"><text style="color: #22a56e;" >- </text></td>
<td style="vertical-align: middle;">67.4</td>
<td style="vertical-align: middle;">77.4</td>
<td style="vertical-align: middle;"><text style="color: #22a56e;" >- </text></td>
</tr>
<tr>
<td style="vertical-align: middle;"></td>
<td style="text-align: left; padding: 2px 10px;"><b class=""><a href="https://huggingface.co/lmms-lab/LLaVA-NeXT-Video-72B-Qwen2" class="ext-link" style="font-size: 16px; margin-left: 5px;">LLaVA-Video
<p style="font-size: 12px; margin-left: 5px; color: #858383;">Bytedance & NTU S-Lab</p>
</a></b></td>
<td style="vertical-align: middle;">72B</td>
<td style="vertical-align: middle;">64 </td>
<td style="vertical-align: middle;"><strong>70.3</strong></td>
<td style="vertical-align: middle;"><strong>75.9</strong></td>
<td style="vertical-align: middle;"> <strong style="color: #22a56e;" >77.4</strong></td>
<td style="vertical-align: middle;">80.1</td>
<td style="vertical-align: middle;">81.8</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >82.8</text></td>
<td style="vertical-align: middle;">68.7</td>
<td style="vertical-align: middle;">73.8</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >76.3</text></td>
<td style="vertical-align: middle;">62.1</td>
<td style="vertical-align: middle;">72.2</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >73.1</text></td>
</tr>
<tr>
<td style="vertical-align: middle;"></td>
<td style="text-align: left; padding: 2px 10px;"><b class=""><a href="https://openai.com/index/hello-gpt-4o/" class="ext-link" style="font-size: 16px; margin-left: 5px;">GPT-4o
<p style="font-size: 12px; margin-left: 5px; color: #858383;">OpenAI</p>
</a></b></td>
<td style="vertical-align: middle;">- </td>
<td style="vertical-align: middle;">384 </td>
<td style="vertical-align: middle;"><b class="">71.9</b></td>
<td style="vertical-align: middle;"><b class="">77.2</b></td>
<td style="vertical-align: middle;"><strong style="color: #22a56e;" >-</strong></td>
<td style="vertical-align: middle;">80.0</td>
<td style="vertical-align: middle;">82.8</td>
<td style="vertical-align: middle;"><text style="color: #22a56e;" >- </text></td>
<td style="vertical-align: middle;">70.3</td>
<td style="vertical-align: middle;">76.6</td>
<td style="vertical-align: middle;"><text style="color: #22a56e;" >- </text></td>
<td style="vertical-align: middle;">65.3</td>
<td style="vertical-align: middle;">72.1</td>
<td style="vertical-align: middle;"><text style="color: #22a56e;" >- </text></td>
</tr>
<tr>
<td style="vertical-align: middle;"></td>
<td style="text-align: left; padding: 2px 10px;"><b class=""><a href="https://github.com/QwenLM/Qwen2-VL" class="ext-link" style="font-size: 16px; margin-left: 5px;">Qwen2-VL
<p style="font-size: 12px; margin-left: 5px; color: #858383;">Alibaba</p>
</a></b></td>
<td style="vertical-align: middle;">72B </td>
<td style="vertical-align: middle;">32 </td>
<td style="vertical-align: middle;"><b class="">64.9</b></td>
<td style="vertical-align: middle;"><b class="">71.9</b></td>
<td style="vertical-align: middle;"> <strong style="color: #22a56e;" >72.9</strong></td>
<td style="vertical-align: middle;">75.0</td>
<td style="vertical-align: middle;">76.7</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >77.4</text></td>
<td style="vertical-align: middle;">63.3</td>
<td style="vertical-align: middle;">69.9</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >70.2</text></td>
<td style="vertical-align: middle;">56.3</td>
<td style="vertical-align: middle;">69.2</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >71.0</text></td>
</tr>
<tr>
<td style="vertical-align: middle;"></td>
<td style="text-align: left; padding: 2px 10px;"><b class=""><a href="https://huggingface.co/aws-prototyping/long-llava-qwen2-7b" class="ext-link" style="font-size: 16px; margin-left: 5px;">Long-LLaVA
<p style="font-size: 12px; margin-left: 5px; color: #858383;">Amazon</p>
</a> </b></td>
<td style="vertical-align: middle;">7B </td>
<td style="vertical-align: middle;">64 </td>
<td style="vertical-align: middle;"><b class="">52.9</b></td>
<td style="vertical-align: middle;"><b class="">57.1</b></td>
<td style="vertical-align: middle;"> <strong style="color: #22a56e;" >62.6</strong></td>
<td style="vertical-align: middle;">61.9 </td>
<td style="vertical-align: middle;">66.2 </td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >67.1</text></td>
<td style="vertical-align: middle;">51.4 </td>
<td style="vertical-align: middle;">54.7 </td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >60.4</text></td>
<td style="vertical-align: middle;">45.4 </td>
<td style="vertical-align: middle;">50.3 </td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >60.1</text></td>
</tr>
<tr>
<td style="vertical-align: middle;"></td>
<td style="text-align: left; padding: 2px 10px;"><b class=""><a href="https://github.com/EvolvingLMMs-Lab/LongVA" class="ext-link" style="font-size: 16px; margin-left: 5px;">LongVA
<p style="font-size: 12px; margin-left: 5px; color: #858383;">NTU S-Lab</p>
</a></b></td>
<td style="vertical-align: middle;">7B</td>
<td style="vertical-align: middle;">128 </td>
<td style="vertical-align: middle;"><strong>52.6</strong></td>
<td style="vertical-align: middle;"><strong>54.3</strong></td>
<td style="vertical-align: middle;"> <strong style="color: #22a56e;" >62.0</strong></td>
<td style="vertical-align: middle;">61.1</td>
<td style="vertical-align: middle;">61.6</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >66.1</text></td>
<td style="vertical-align: middle;">50.4</td>
<td style="vertical-align: middle;">53.6</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >60.4</text></td>
<td style="vertical-align: middle;">46.2</td>
<td style="vertical-align: middle;">47.6</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >59.4</text></td>
</tr>
<tr>
<td style="vertical-align: middle;"></td>
<td style="text-align: left; padding: 2px 10px;"><b class=""><a href="https://llava-vl.github.io/blog/2024-04-30-llava-next-video/" class="ext-link" style="font-size: 16px; margin-left: 5px;">LLaVA-NeXT-Video
<p style="font-size: 12px; margin-left: 5px; color: #858383;">NTU</p>
</a></b></td>
<td style="vertical-align: middle;">7B</td>
<td style="vertical-align: middle;">16 </td>
<td style="vertical-align: middle;"><strong>43.0</strong></td>
<td style="vertical-align: middle;"><strong>47.7</strong></td>
<td style="vertical-align: middle;"> <strong style="color: #22a56e;" >50.0</strong></td>
<td style="vertical-align: middle;">49.4</td>
<td style="vertical-align: middle;">51.8</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >56.6</text></td>
<td style="vertical-align: middle;">43.0</td>
<td style="vertical-align: middle;">46.4</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >47.4</text></td>
<td style="vertical-align: middle;">36.7</td>
<td style="vertical-align: middle;">44.9</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >46.0</text></td>
</tr>
<tr>
<td style="vertical-align: middle;"></td>
<td style="text-align: left; padding: 2px 10px;"><b class=""><a href="https://github.com/PKU-YuanGroup/Video-LLaVA" class="ext-link" style="font-size: 16px; margin-left: 5px;">Video-LLaVA
<p style="font-size: 12px; margin-left: 5px; color: #858383;">PKU</p>
</a></b></td>
<td style="vertical-align: middle;">7B</td>
<td style="vertical-align: middle;">8 </td>
<td style="vertical-align: middle;"><strong>39.9</strong></td>
<td style="vertical-align: middle;"><strong>41.6</strong></td>
<td style="vertical-align: middle;"> <strong style="color: #22a56e;" >45.0</strong></td>
<td style="vertical-align: middle;">45.3</td>
<td style="vertical-align: middle;">46.1</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >49.5</text></td>
<td style="vertical-align: middle;">38.0</td>
<td style="vertical-align: middle;">40.7</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >43.0</text></td>
<td style="vertical-align: middle;">36.2</td>
<td style="vertical-align: middle;">38.1</td>
<td style="vertical-align: middle;"> <text style="color: #22a56e;" >42.5</text></td>
</tr>
</tbody>
</table>
</div>
</div>
</div>
</div>
</section>
<!-- DATASET SECTION -->
<section class="hero is-light is-small">
<div class="hero-body has-text-centered">
<h1 class="title is-1 mathvista_other">
<span class="mathvista_other" style="vertical-align: middle">Video-RAG</span>
</h1>
</div>
</section>
<section class="section">
<div class="container">
<div class="columns is-centered">
<div class="column">
<div class="content has-text-centered">
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3">Highlights</h2>
<div class="box m-5">
<div class="content has-text-centered">
<div style="display: flex; justify-content: center;">
<img src="static/images/radar.png" alt="grade-lv" style="width: 100%;"/>
</div>
<p style="text-align: center; margin-left: auto; margin-right: auto; width: 95%;">
Comparison of the performance of Video-RAG with LLaVA-Video-72B, Gemini-1.5-Pro, and GPT-4o across various benchmarks, including the sub-tasks from Video-MME (here we focus only on those that outperform Gemini-1.5-Pro), LongVideoBench, and MLVU benchmarks.</p>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths" style="width: 100%;">
<h2 class="title is-3">Examples</h2>
<div class="content has-text-justified">
<p style="text-align: center;", class="mt-3"><strong>We apply Video-RAG on LLaVA-Video for visualization.</strong>
</p>
<div id="results-carousel" class="carousel results-carousel">
<div class="content has-text-centered">
<img src="static/images/highlight_video/hl1.png" style="width: 95%;"/>
<p style="margin-bottom: 30px;">🔗 Full Video Link: <a href="https://www.youtube.com/watch?v=fFjv93ACGo8">https://www.youtube.com/watch?v=fFjv93ACGo8</a></p>
</div>
<div class="content has-text-centered">
<img src="static/images/highlight_video/hl2.png" style="width: 95%;"/>
<p style="margin-bottom: 30px;">🔗 Full Video Link: <a href="https://www.youtube.com/watch?v=3m29MQ-qPfg">https://www.youtube.com/watch?v=3m29MQ-qPfg</a></p>
</div>
<div class="content has-text-centered">
<img src="static/images/highlight_video/hl3.png" style="width: 95%;"/>
<p style="margin-bottom: 30px;">🔗 Full Video Link: <a href="https://www.youtube.com/watch?v=QHFy-nWNJYk">https://www.youtube.com/watch?v=QHFy-nWNJYk</a></p>
</div>
<div class="content has-text-centered">
<img src="static/images/highlight_video/hl4.png" style="width: 95%;"/>
<p style="margin-bottom: 30px;">🔗 Full Video Link: <a href="https://www.youtube.com/watch?v=9HPFkUhOp1Q">https://www.youtube.com/watch?v=9HPFkUhOp1Q</a></p>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
<div class="columns is-centered">
<div class="column">
<div class="content has-text-centered">
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3">Framework of Video-RAG</h2>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/framework.png" alt="grade-lv" width="100%"/>
<p style="text-align: center; margin-left: auto; margin-right: auto; width: 95%;">
In the query decouple phase, the LVLM is prompted to generate a retrieval request for auxiliary texts. Next, in the auxiliary text generation and retrieval phase, the video is processed in parallel to extract three types of textual information (OCR, ASR, and object detection), and the relevant text is retrieved as the auxiliary text. Finally, in the integration and generation phase, auxiliary texts are combined with the query and the video to generate the response.</p>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
<div class="columns is-centered">
<div class="column">
<div class="content has-text-centered">
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3">Visually-aligned Visulization</h2>
<div class="box m-5">
<div class="content has-text-centered">
<div style="display: flex; justify-content: center;">
<img src="static/images/visually-aligned.png" alt="grade-lv" style="width: 100%;"/>
</div>
<p style="text-align: center; margin-left: auto; margin-right: auto; width: 95%;">
Grad-CAM visualizations of the last hidden state heatmap along with t-SNE visualizations of the user's query and keyframe features of the first example shown in Examples section.
As demonstrated, the retrieved auxiliary texts help <strong>cross-modality alignment</strong> by assisting the model to <strong>pay more attention to query-relevant keyframes</strong> and thus generate more robust and accurate answers to the user's query.
</p>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
</section>
<section class="hero is-light is-small">
<div class="hero-body has-text-centered">
<h1 class="title is-1 mathvista_other" id="citation">Citation</h1>
</div>
</section>
<section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<pre><code>@misc{luo2024videoragvisuallyalignedretrievalaugmentedlong,
title={Video-RAG: Visually-aligned Retrieval-Augmented Long Video Comprehension},
author={Yongdong Luo and Xiawu Zheng and Xiao Yang and Guilin Li and Haojia Lin and Jinfa Huang and Jiayi Ji and Fei Chao and Jiebo Luo and Rongrong Ji},
year={2024},
eprint={2411.13093},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2411.13093},
}
</code></pre>
</div>
</section>
<section class="section">
<div class="container" style="width: 60%;">
<style>
pre {
background-color: #f4f4f4;
padding: 5px; /* 调整padding为5px */
border: 1px solid #ddd;
border-radius: 5px;
overflow-x: auto; /* 允许水平滚动 */
}
code {
font-family: Consolas, "Courier New", monospace;
color: #d63384; /* 代码文本颜色 */
}
</style>
</div>
</section>
<footer class="footer">
<!-- <div class="container"> -->
<div class="content has-text-centered">
</div>
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p style="text-align: center;">
This website is adapted from <a href="https://github.com/BradyFU/Video-MME">Video-MME</a>.
</p>
<p style="text-align: center;">
Feel free to contact us at yongdongluo1207 AT 163.com
</p>
</div>
</div>
</div>
<!-- </div> -->
</footer>
</body>
</html>