-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
859 lines (809 loc) · 39.1 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
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
<!DOCTYPE html>
<html>
<head>
<title>VidEgoThink</title>
<style>
.hidden {
display: none;
}
</style>
<script src="https://cdn.jsdelivr.net/npm/chart.js"></script>
<meta charset="utf-8">
<meta name="description"
content="Assessing Egocentric Video Understanding Capabilities for Embodied AI">
<meta name="keywords" content="VidEgoThink, Egocentric, Embodied AI, Vision Language Model, First-Person, Benchmark, Artificial Intelligence, AI, AGI">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title> VidEgoThink: Assessing Egocentric Video Understanding Capabilities for Embodied AI</title>
<link rel="icon" href="./static/images/VidEgoThink_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="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="./static/css/index.css">
<link rel="stylesheet" href="./static/css/leaderboard.css">
<!-- <link href="https://unpkg.com/tabulator-tables@5.5.2/dist/css/tabulator_bulma.min.css" rel="stylesheet">
<script type="text/javascript" src="https://unpkg.com/tabulator-tables@5.5.2/dist/js/tabulator.min.js"></script> -->
<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/index.js"></script>
<script src="./static/js/question_card.js"></script>
<script src="./data/results/data_setting.js" defer></script>
<script src="./data/results/model_scores.js" defer></script>
<script src="./visualizer/data/data_public.js" defer></script>
</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">
<a class="navbar-link">
More Research
</a>
<div class="navbar-dropdown">
<!-- <a class="navbar-item" href="https://github.com/imoneoi/openchat">
<b>OpenChat</b> <p style="font-size:18px; display: inline; margin-left: 5px;">🔥</p>
</a> -->
<a class="navbar-item" href="https://adacheng.github.io/EgoThink/">
<b>EgoThink</b>
</a>
<a class="navbar-item" href="https://zhichengg.github.io/stb.github.io/">
<b>StableToolBench</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">
<img src="static/images/VidEgoThink_icon.png" style="width:1em;vertical-align: middle" alt="Logo"/>
<span class="mmmu" style="vertical-align: middle">VidEgoThink</span>
</h1>
<h2 class="subtitle is-3 publication-subtitle">
Assessing Egocentric Video Understanding Capabilities for Embodied AI
</h2>
<div class="is-size-5 publication-authors">
<span class="author-block">Sijie Cheng<sup style="color:#6fbf73;">†,1,2,6</sup>,</span>
<span class="author-block">Kechen Fang*<sup style="color:#ffac33;">2,5</sup>,</span>
<span class="author-block">Yangyang Yu*<sup style="color:#ed4b82;">2,5</sup>,</span>
<span class="author-block">Sicheng Zhou*<sup style="color:#007bff;">2,3</sup>,</span><br>
<span class="author-block">Bohao Li<sup style="color:#ed4b82;">4,6</sup>,</span>
<span class="author-block">Ye Tian<sup style="color:#9b51e0;">6</sup>,</span>
<span class="author-block">Tingguang Li<sup style="color:#9b51e0;">6</sup>,</span>
<span class="author-block">Lei Han<sup style="color:#9b51e0;">✉,6</sup>,</span>
<span class="author-block">Yang Liu<sup style="color:#ed4b82;">✉,1,2</sup>,</span>
</div>
<br>
<div class="is-size-5 publication-authors">
<span class="author-block"><sup style="color:#6fbf73;">1</sup>Department of Computer Science and Technology, Tsinghua University</span><br>
<span class="author-block"><sup style="color:#ffac33;">2</sup>Institute for AI Industry Research (AIR), Tsinghua University</span><br>
<span class="author-block"><sup style="color:#ed4b82;">3</sup>Department of Mechanical and Industrial Engineering, University of Toronto</span><br>
<span class="author-block"><sup style="color:#007bff;">4</sup>School of Data Science, The Chinese University of HongKong</span></br>
<span class="author-block"><sup style="color:#ffac33;">5</sup>Zhili College, Tsinghua University</span>
<span class="author-block"><sup style="color:#9b51e0;">6</sup>Tencent Robotics X</span><br>
<!-- <span class="author-block"><sup style="color:#ed4b82;">6</sup>University of Victoria,</span>
<span class="author-block"><sup style="color:#9b51e0;">7</sup>Princeton University</span> -->
</div>
<br>
<div class="is-size-5 publication-authors">
<span class="author-block">*Equal contribution, ✉Corresponding author</span><br>
<span class="author-block">†Project Lead:</span>
<span class="author-block"><a href="mailto:csj23@mails.tsinghua.edu.cn">csj23@mails.tsinghua.edu.cn</a></span>
<!-- <span class="author-block"><a href="mailto:su.809@osu.edu">su.809@osu.edu</a>,</span>
<span class="author-block"><a href="mailto:wenhuchen@uwaterloo.ca">wenhuchen@uwaterloo.ca</a></span> -->
</div>
<div class="column has-text-centered">
<div class="publication-links">
<!-- PDF Link. -->
<span class="link-block">
<a href="https://arxiv.org/pdf/2410.11623"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>arXiv</span>
</a>
</span>
<span class="link-block">
<a href="https://huggingface.co/papers/2410.11623"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<p style="font-size:18px">🤗</p>
</span>
<span>HF Paper</span>
</a>
</span>
<!-- <span class="link-block"> -->
<!-- @TODO: change links -->
<!-- <a href="https://huggingface.co/datasets/" -->
<!-- class="external-link button is-normal is-rounded is-dark"> -->
<!-- <span class="icon"> -->
<!-- <i class="far fa-images"></i> -->
<!-- <p style="font-size:18px">🤗</p> -->
<!-- 🔗 -->
<!-- </span> -->
<!-- <span>Dataset</span> -->
<!-- </a> -->
<!-- </span> -->
<span class="link-block">
<a href="https://github.com/AdaCheng/VidEgoThink"
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>
<!-- Dataset Link. -->
<!-- Leaderboard Link. -->
<span class="link-block">
<a href="#leaderboard"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<p style="font-size:18px">🏆</p>
</span>
<span>Leaderboard</span>
</a>
</span>
<!-- Visualization Link. -->
<!-- <span class="link-block">
<a href="#examples"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<p style="font-size:18px">📖</p>
</span>
<span>Examples</span>
</a>
</span> -->
<!-- Twitter Link. -->
<!-- <span class="link-block">
<a href="https://twitter.com/xiangyue96/status/1729698316554801358"
class="external-link button is-normal is-rounded is-dark">
<span class="icon"> -->
<!-- <i class="far fa-images"></i> -->
<!-- 💻🔗 -->
<!-- <p style="font-size:18px">🌐</p> -->
<!-- </span>
<span>Twitter</span>
</a>
</span> -->
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<style>
.center {
display: block;
margin-left: auto;
margin-right: auto;
width: 80%;
}
</style>
<section class="hero teaser">
<div class="container is-max-desktop">
<!-- <div class="hero-body">
<img src="static/images/tease_scores.png" alt="Examples from the dataset"/>
<h2 class="subtitle has-text-centered">
<span class="dnerf">Nerfies</span> turns selfie videos from your phone into
free-viewpoint
portraits.
</h2>
</div> -->
<!-- <div class="box m-5"> -->
<div class="content has-text-centered">
<img src="static/images/intro.png" alt="geometric reasoning" />
<p> <b><i>Figure 1:</i></b> The main tasks of VidEgoThink benchmark to comprehensively assess the egocentric video understanding capabilities in Embodied AI. There are four types of tasks, including video question answering, hierarchy planning, visual grounding, and reward modeling. These four tasks are complementary to each other to implement a complete goal for Embodied AI. </p>
</div>
<!-- </div> -->
<!-- </div> -->
</div>
</section>
<section class="section">
<div class="container" style="margin-bottom: 2vh;">
<!-- Abstract. -->
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">🔔News</h2>
<div class="content has-text-justified">
<p>
<b>[2024-10]: <a href="https://huggingface.co/papers/2410.11623">VidEgoThink</a> is the Top-1 paper of Oct-17 in Hugging Face. 🔥</b><br>
<b>[2024-10]: Our paper <a href="https://arxiv.org/abs/2410.11623">VidEgoThink: Assessing Egocentric Video Understanding Capabilities for Embodied AI</a> has been released.</b><br>
<b>[2024-09]: EgoThink and VidEgoThink is invited to be presented in <a href="https://zhidx.com/p/441426.html">ZhiDX</a>.</b><br>
</p>
</div>
<h2 class="title is-3">Abstraction</h2>
<div class="content has-text-justified">
<p>
Recent advancements in Multi-modal Large Language Models (MLLMs) have opened new avenues for applications in Embodied AI. Building on previous work, EgoThink, we introduce VidEgoThink, a comprehensive benchmark for evaluating egocentric video understanding capabilities. To bridge the gap between MLLMs and low-level control in Embodied AI, we design four key interrelated tasks: video question-answering, hierarchy planning, visual grounding and reward modeling. To minimize manual annotation costs, we develop an automatic data generation pipeline based on the Ego4D dataset, leveraging the prior knowledge and multimodal capabilities of GPT-4o. Three human annotators then filter the generated data to ensure diversity and quality, resulting in the VidEgoThink benchmark. We conduct extensive experiments with three types of models: API-based MLLMs, open-source image-based MLLMs, and open-source video-based MLLMs. Experimental results indicate that all MLLMs, including GPT-4o, perform poorly across all tasks related to egocentric video understanding. These findings suggest that foundation models still require significant advancements to be effectively applied to first-person scenarios in Embodied AI. In conclusion, VidEgoThink reflects a research trend towards employing MLLMs for egocentric vision, akin to human capabilities, enabling active observation and interaction in the complex real-world environments.
</p>
</div>
</div>
</div>
<!--/ Abstract. -->
</div>
</section>
<!-- DATASET SECTION -->
<section class="hero is-light is-small">
<div class="hero-body has-text-centered">
<h1 class="title is-1 mmmu">
<!-- <img src="static/images/VidEgoThink_icon.png" style="width:1em;vertical-align: middle" alt="Logo"/> -->
<span class="mmmu" style="vertical-align: middle">Background</span>
</h1>
</div>
</section>
<section class="section">
<div class="container">
<div class="columns is-centered has-text-centered">
<!-- <div class="column is-full-width has-text-centered"> -->
<div class="column is-four-fifths">
<!-- <h2 class="title is-3">Comparisons with Existing Benchmarks</h2> -->
<div class="content has-text-justified">
<p>
Various egocentric benchmarks have emerged to evaluate the capabilities of MLLMs from a first-person perspective. However, the absence of a <b>comprehensive</b> video benchmark from the egocentric perspective presents a significant challenge to the development of general foundation models. Furthermore, current benchmarks, both in task design and textual output forms, focus on traditional video question-answering settings and neglect the potential to support downstream applications in Embodied AI, such as glass devices or autonomous robots. Therefore, it is crucial to design <b>suitable task formats that can be effectively applied to downstream applications in Embodied AI</b>.
</p>
<div class="content has-text-centered">
<img src="static/images/comparison.png" alt="algebraic reasoning" class="center">
<p><b><i>Table 1:</i></b> omparison of recent evaluation benchmarks of multimodal large language models and our proposed benchmark VidEgoThink. VQA/HP/VG/RM indicate visual question answering, hierarchy planning, visual grounding, and reward modeling. Existing/Handcraft/Automatic denote the way of collecting data, including existing dataset, manual annotation, and automatic generation.</p>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="hero is-light is-small">
<div class="hero-body has-text-centered">
<h1 class="title is-1 mmmu">
<img src="static/images/VidEgoThink_icon.png" style="width:1em;vertical-align: middle" alt="Logo"/>
<span class="mmmu" style="vertical-align: middle">VidEgoThink Benchmark</span>
</h1>
</div>
</section>
<section class="section">
<div class="container">
<div class="columns is-centered has-text-centered">
<!-- <div class="column is-full-width has-text-centered"> -->
<div class="column is-four-fifths">
<div class="content has-text-justified">
<p>
Given that the utilization of foundation models in Embodied AI remains an open research question, we carefully design four types of interrelated tasks for comprehensive assessment: (i) video question-answering, (ii) hierarchy planning, (iii) visual grounding, (iv) reward modeling.
</p>
</div>
</div>
</div>
<div class="columns is-centered has-text-centered">
<!-- <div class="column is-full-width has-text-centered"> -->
<div class="column is-four-fifths">
<h2 class="title is-3">1. Video Question Answering</h2>
<div class="content has-text-justified">
<p>
Previous evaluation studies on egocentric vision have predominantly focused on static images, constrained by the input format limitations of earlier MLLMs. However, recent advancements in API-based and video-based MLLMs have demonstrated significant progress. Since our real world is inherently dynamic and humans frequently process substantial amounts of video data, it is crucial to evaluate the video understanding capabilities of MLLMs. Considering the essential abilities for observing and interacting with the real world from a first-person perspective, we decompose the content of video modalities around “myself” into three main elements: object, action, and scene. Furthermore, we explore a series of fine-grained dimensions from these elements.
</p>
<img src="static/images/vqa.png" alt="algebraic reasoning" class="center">
<div class="content has-text-centered">
<p><b><i>Figure 2:</i></b> Case of video question answering.</p>
</div>
</div>
</div>
</div>
<div class="columns is-centered has-text-centered">
<!-- <div class="column is-full-width has-text-centered"> -->
<div class="column is-four-fifths">
<h2 class="title is-3">2. Hierarchy Planning</h2>
<div class="content has-text-justified">
<p>
Recently, a hierarchy planning framework has been proposed to combine the advantages of foundation models and traditional methods in Embodied AI. In detail, foundation models are used as the planner to decompose high-level task instructions (e.g., “cook salmon”) into either mid-level steps (e.g., “# put salmon in the microwave’) or low-level atomic actions (e.g., “find(microwave)”), which is much more convenient for controlling. Therefore, we design two types of planning tasks: high-level goal to mid-level step (High-to-Mid), and mid-level step to low-level action (Mid-to-Low).
</p>
<img src="static/images/hp.png" alt="algebraic reasoning" class="center">
<div class="content has-text-centered">
<p><b><i>Figure 3:</i></b> Case of hierarchy planning.</p>
</div>
</div>
</div>
</div>
<div class="columns is-centered has-text-centered">
<!-- <div class="column is-full-width has-text-centered"> -->
<div class="column is-four-fifths">
<h2 class="title is-3">3. Visual Grounding</h2>
<div class="content has-text-justified">
<p>
While natural language is effective for human communication, it cannot be directly translated into low-level actions or grounded in the real world. Consequently, visual grounding has garnered significant attention in both image- and video-based MLLMs. This task requires models to ground complex natural language descriptions or instructions in an image or video and output the corresponding pixel-level bounding boxes, masks, or frames. The bounding boxes and masks can directly identify actionable objects, while the frames can provide sufficient spatial or temporal information for downstream tasks. Therefore, we specifically design three tasks for different situations: object grounding, frame grounding, and temporal grounding.
</p>
<img src="static/images/vg.png" alt="algebraic reasoning" class="center">
<div class="content has-text-centered">
<p><b><i>Figure 4:</i></b> Case of visual grounding.</p>
</div>
</div>
</div>
</div>
<div class="columns is-centered has-text-centered">
<!-- <div class="column is-full-width has-text-centered"> -->
<div class="column is-four-fifths">
<h2 class="title is-3">4. Reward Modeling</h2>
<div class="content has-text-justified">
<p>
In Embodied AI, manually designing reward functions to supervise actions is challenging due to the need for accuracy and diversity, especially for human activities. Benefiting from the large-scale Internet training corpus, foundation models can serve as reward models with built-in commonsense and reasoning capabilities. As a reward model, MLLMs should first observe the video to determine the completion status of the target motion. If the action is not completed, the reward model should further provide fine-grained feedback to help achieve the goal. Hence, we specifically design two types of tasks: critique, and feedback.
</p>
<img src="static/images/rm.png" alt="algebraic reasoning" class="center">
<div class="content has-text-centered">
<p><b><i>Figure 5:</i></b> Case of reward modeling.</p>
</div>
</div>
</div>
</div>
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3">Statistics</h2>
<!-- <div id="results-carousel" class="carousel results-carousel"> -->
<div class="content has-text-centered">
<img src="static/images/statistics.png" alt="algebraic reasoning" width="75%"/>
<p><b><i>Table 2:</i></b> The statistics of videos across different benchmarks. Duration denotes the average time duration in second of all videos. LenQ and LenA indicate that the average length of questions and answers in the word level. TypeQ denotes the type of questions.</p>
</div>
</div>
</div>
</div>
</section>
<!-- RESULTS SECTION -->
<section class="hero is-light is-small">
<div class="hero-body has-text-centered">
<h1 class="title is-1 mmmu" id="leaderboard">Leaderboard</h1>
</div>
</section>
<!-------------------------------------------------------------------- RESULTS SECTION -------------------------------------------------------------------->
<section class="section">
<div class="container">
<!-- <div class="columns is-centered has-text-centered"> -->
<!-- <div class="column is-full-width has-text-centered"> -->
<!-- <div class="column is-four-fifths"> -->
<!-- <h2 class="title is-3">Vision-Language Models</h2> -->
<!-- <div class="content has-text-centered"> -->
<!-- <img src="static/images/vlms.png" alt="algebraic reasoning" class="center"> -->
<!-- <p><b><i>Table 3:</i></b> Statistics of compared API-based and open-source VLMs, where TTP and ToP indicate Total Trainable Parameters and Total Parameters, respectively. Moreover, EgoData and Video indicate that there are egocentric visual data and video data for training, respectively.</p> -->
<!-- </div> -->
<!-- </div> -->
<!-- </div> -->
<div class="columns is-centered has-text-centered">
<!-- <div class="column is-full-width has-text-centered"> -->
<div class="column is-four-fifths">
<h2 class="title is-3">Video Question Answering</h2>
<div class="content has-text-justified">
<img src="static/images/result_vqa.png" alt="algebraic reasoning" class="center">
<div class="content has-text-centered">
<p><b><i>Table 3:</i></b> Experimental results of video question answering. OE, OO, OI, OC, OS, OP denote object existence, object order, object interaction, object count, object state, object prediction. AE, AS, AC indicates action existence, action sequence, action count. SE, ST, SP denote scene existence, scene transition, scene prediction. The bold font denotes the best performance and the underline font denotes the second-best performance.</p>
</div>
</div>
</div>
</div>
<div class="columns is-centered has-text-centered">
<!-- <div class="column is-full-width has-text-centered"> -->
<div class="column is-four-fifths">
<h2 class="title is-3">All Tasks</h2>
<div class="content has-text-justified">
<img src="static/images/result_all.png" alt="algebraic reasoning" class="center">
<div class="content has-text-centered">
<p><b><i>Table 4:</i></b> Experimental results of video question answerng, hierarchy planning, visual grounding, and reward modeling tasks. The bold font denotes the best performance and the underline font denotes the second-best performance.</p>
</div>
</div>
</div>
</div>
<!-- <div class="columns is-centered m-6"> -->
<!-- <div class="column is-full has-text-centered content"> -->
<!-- <h2 class="title is-3" id="leaderboard">Leaderboard</h2> -->
<!-- <div class="content"> -->
<!-- <div class="content has-text-justified"> -->
<!-- <p> -->
<!-- Evaluating open-ended model generations is a non-trivial problem. To address this, we propose using GPT-4 as an automatic evaluator to better measure the generated answers. We continuously update the results of recent VLMs to ensure the effectiveness of EgoThink. Feel free to contribute to the performance of your model by adding it to our <a href="https://github.com/AdaCheng/EgoThink/blob/main/index.html">index.html</a>; we will review and merge it accordingly. -->
<!-- </p> -->
<!-- </div> -->
<!-- <button id="toggleButton" onclick="changeButtonText()">Switch to Validation Set</button> -->
<!-- <table id="table1" class="js-sort-table">
<tr>
<td class="js-sort-number"><strong>Reset</strong></td>
<td class="js-sort-number"><strong>Average</strong></td>
<td class="js-sort-number"><strong>Exist</strong></td>
<td class="js-sort-number"><strong>Attr</strong></td>
<td class="js-sort-number"><strong>Afford</strong></td>
<td class="js-sort-number"><strong>Activity</strong></td>
<td class="js-sort-number"><strong>Loc</strong></td>
<td class="js-sort-number"><strong>Spatial</strong></td>
<td class="js-sort-number"><strong>Count</strong></td>
<td class="js-sort-number"><strong>Compar</strong></td>
<td class="js-sort-number"><strong>Situtaed</strong></td>
<td class="js-sort-number"><strong>Forecasting</strong></td>
<td class="js-sort-number"><strong>Nav</strong></td>
<td class="js-sort-number"><strong>Assist</strong></td>
</tr>
<tr style="background-color: #f8fffe;">
<td style="text-align: left;"><b>GPT-4V(ision)</b></td>
<td><b>65.5</b></td>
<td>62.0</td>
<td><b>82.0</b></td>
<td><b>58.0</b></td>
<td><b>59.5</b></td>
<td style="text-decoration: underline;">86.0</td>
<td style="text-decoration: underline;">62.0</td>
<td><b>42.0</b></td>
<td>48.0</td>
<td><b>83.0</b></td>
<td><b>55.0</b></td>
<td><b>64.0</b></td>
<td><b>84.0</b></td>
</tr>
<tr style="background-color: #f9f2f8;">
<td style="text-align: left;"><b>OpenFlamingo-7B</b></td>
<td>27.2</td>
<td>16.0</td>
<td>55.0</td>
<td>37.0</td>
<td>15.0</td>
<td>34.0</td>
<td>34.0</td>
<td>21.0</td>
<td>40.0</td>
<td>21.0</td>
<td>31.0</td>
<td>11.0</td>
<td>11.0</td>
</tr>
<tr style="background-color: #f9f2f8;">
<td style="text-align: left;"><b>BLIP-2-6.7B</b></td>
<td>28.1</td>
<td>49.0</td>
<td>29.0</td>
<td>39.0</td>
<td>33.5</td>
<td>60.0</td>
<td>31.0</td>
<td>3.0</td>
<td>21.0</td>
<td>33.0</td>
<td>25.0</td>
<td>8.0</td>
<td>6.0</td>
</tr>
<tr style="background-color: #f9f2f8;">
<td style="text-align: left;"><b>LLaVA-1.5-7B</b></td>
<td>39.0</td>
<td>33.0</td>
<td>47.0</td>
<td style="text-decoration: underline;">54.0</td>
<td>35.5</td>
<td>35.0</td>
<td>49.0</td>
<td>20.0</td>
<td>47.0</td>
<td>37.0</td>
<td>27.0</td>
<td>29.0</td>
<td>54.0</td>
</tr>
<tr style="background-color: #f9f2f8;">
<td style="text-align: left;"><b>MiniGPT-4-7B</b></td>
<td>40.6</td>
<td>50.0</td>
<td>56.0</td>
<td>46.0</td>
<td>39.0</td>
<td>55.0</td>
<td>49.0</td>
<td>14.0</td>
<td>48.0</td>
<td>31.0</td>
<td>41.5</td>
<td>14.0</td>
<td>44.0</td>
</tr>
<tr style="background-color: #f9f2f8;">
<td style="text-align: left;"><b>InstructBLIP-7B</b></td>
<td>42.4</td>
<td>50.0</td>
<td>33.0</td>
<td>45.0</td>
<td>47.5</td>
<td>77.0</td>
<td>38.0</td>
<td>18.0</td>
<td>43.0</td>
<td>67.0</td>
<td>40.5</td>
<td>19.0</td>
<td>31.0</td>
</tr>
<tr style="background-color: #f9f2f8;">
<td style="text-align: left;"><b>LLaMA-Adapter-7B</b></td>
<td>42.5</td>
<td>37.0</td>
<td>60.0</td>
<td>46.0</td>
<td>34.5</td>
<td>48.0</td>
<td>51.0</td>
<td>29.0</td>
<td>39.0</td>
<td>25.0</td>
<td>41.5</td>
<td>42.0</td>
<td>57.0</td>
</tr>
<tr style="background-color: #f9f2f8;">
<td style="text-align: left;"><b>Otter-I-7B</b></td>
<td>45.3</td>
<td>48.0</td>
<td>56.0</td>
<td>39.0</td>
<td>44.0</td>
<td>60.0</td>
<td>44.0</td>
<td>39.0</td>
<td>48.0</td>
<td>42.0</td>
<td>38.0</td>
<td>31.0</td>
<td>55.0</td>
</tr>
<tr style="background-color: #f9f2f8;">
<td style="text-align: left;"><b>PandaGPT-7B</b></td>
<td>46.2</td>
<td>40.0</td>
<td>56.0</td>
<td>41.0</td>
<td>37.0</td>
<td>61.0</td>
<td>52.0</td>
<td>19.0</td>
<td style="text-decoration: underline;">52.0</td>
<td>53.0</td>
<td>43.0</td>
<td>39.0</td>
<td>61.0</td>
</tr>
<tr style="background-color: #f9f2f8;">
<td style="text-align: left;"><b>mPLUG-owl-7B</b></td>
<td>48.8</td>
<td>56.0</td>
<td>58.0</td>
<td>47.0</td>
<td>53.0</td>
<td>60.0</td>
<td>53.0</td>
<td>25.0</td>
<td>49.0</td>
<td>44.0</td>
<td>49.5</td>
<td>33.0</td>
<td>58.0</td>
</tr>
<tr style="background-color: #f9f2f8;">
<td style="text-align: left;"><b>LLaVA-7B</b></td>
<td>49.6</td>
<td>63.0</td>
<td>58.0</td>
<td>50.0</td>
<td>47.0</td>
<td>81.0</td>
<td>45.0</td>
<td>24.0</td>
<td>36.0</td>
<td>47.0</td>
<td>49.5</td>
<td>35.0</td>
<td>60.0</td>
</tr>
<tr style="background-color: #f4f9fe;">
<td style="text-align: left;"><b>InstructBLIP-13B</b></td>
<td>42.8</td>
<td>52.0</td>
<td>55.0</td>
<td>49.0</td>
<td>54.0</td>
<td>63.0</td>
<td>49.0</td>
<td>11.0</td>
<td>33.0</td>
<td>59.0</td>
<td>44.0</td>
<td>19.0</td>
<td>25.0</td>
</tr>
<tr style="background-color: #f4f9fe;">
<td style="text-align: left;"><b>PandaGPT-13B</b></td>
<td>43.1</td>
<td>35.0</td>
<td>52.0</td>
<td>41.0</td>
<td>40.5</td>
<td>68.0</td>
<td>31.0</td>
<td>32.0</td>
<td>40.0</td>
<td>47.0</td>
<td>45.5</td>
<td>16.0</td>
<td>69.0</td>
</tr>
<tr style="background-color: #f4f9fe;">
<td style="text-align: left;"><b>LLaVA-13B-Vicuna</b></td>
<td>46.4</td>
<td>54.0</td>
<td>62.0</td>
<td>52.0</td>
<td>46.0</td>
<td>53.0</td>
<td>46.0</td>
<td>26.0</td>
<td>44.0</td>
<td>29.0</td>
<td>44.0</td>
<td>35.0</td>
<td>66.0</td>
</tr>
<tr style="background-color: #f4f9fe;">
<td style="text-align: left;"><b>BLIP-2-11B</b></td>
<td>49.6</td>
<td>52.0</td>
<td>62.0</td>
<td>41.0</td>
<td>49.5</td>
<td><b>90.0</b></td>
<td><b>66.0</b></td>
<td>25.0</td>
<td>50.0</td>
<td>70.0</td>
<td>48.0</td>
<td>18.0</td>
<td>24.0</td>
</tr>
<tr style="background-color: #f4f9fe;">
<td style="text-align: left;"><b>InstructBLIP-11B</b></td>
<td>51.1</td>
<td><b>74.0</b></td>
<td style="text-decoration: underline;">68.0</td>
<td>48.0</td>
<td>49.5</td>
<td style="text-decoration: underline;">86.0</td>
<td>52.0</td>
<td>32.0</td>
<td>49.0</td>
<td style="text-decoration: underline;">73.0</td>
<td style="text-decoration: underline;">53.0</td>
<td>16.0</td>
<td>17.0</td>
</tr>
<tr style="background-color: #f4f9fe;">
<td style="text-align: left;"><b>LLaVA-13B-Llama2</b></td>
<td>55.1</td>
<td>65.0</td>
<td>61.0</td>
<td>45.0</td>
<td style="text-decoration: underline;">56.0</td>
<td>77.0</td>
<td>53.0</td>
<td style="text-decoration: underline;">34.0</td>
<td>34.0</td>
<td>66.0</td>
<td>50.5</td>
<td style="text-decoration: underline;">49.0</td>
<td style="text-decoration: underline;">71.0</td>
</tr>
<tr style="background-color: #f4f9fe;">
<td style="text-align: left;"><b>LLaVA-1.5-13B</b></td>
<td style="text-decoration: underline;">55.3</td>
<td style="text-decoration: underline;">66.0</td>
<td>55.0</td>
<td>51.0</td>
<td>55.0</td>
<td>82.0</td>
<td>57.0</td>
<td>32.0</td>
<td><b>56.0</b></td>
<td>67.0</td>
<td>48.5</td>
<td>39.0</td>
<td>55.0</td>
</tr>
</table>
<p> <b><i>Table 4:</i></b> Combined single-answer grading scores on zero-shot setups for various dimensions. The <b>bold</b> indicates the best performance while the <u>underline</u> indicates the second-best performance. Exist, Attr, Afford, Loc, Spatial, Count, Compar, Situated, Nav and Assist represent existence, attribute, affordance, location, spatial relationship, counting, comparison, situated reasoning, navigation, and assistance.</p> -->
<!-- </div> -->
<!-- </div> -->
</div>
</section>
<!-------------------------------------------------------------------- RESULTS SECTION -------------------------------------------------------------------->
<!-- @PAN TODO: bibtex -->
<section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title is-3 has-text-centered">BibTeX</h2>
<pre><code>
@article{cheng2024videgothink,
title={VidEgoThink: Assessing Egocentric Video Understanding Capabilities for Embodied AI},
author={Cheng, Sijie and Fang, Kechen and Yu, Yangyang and Zhou, Sicheng and Li, Bohao and Tian, Ye and Li, Tingguang and Han, Lei and Liu, Yang},
journal={arXiv preprint arXiv:2410.11623},
year={2024}
}
</code></pre>
</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 has-text-centered">
<p>
This website is website adapted from <a href="https://adacheng.github.io/EgoThink/">EgoThink</a>, licensed under a <a rel="license"
href="http://creativecommons.org/licenses/by-sa/4.0/">Creative
Commons Attribution-ShareAlike 4.0 International License</a>.
</p>
</div>
</div>
</div>
<!-- </div> -->
</footer>
<style>
.hidden {
display: none;
}
.sortable:hover {
cursor: pointer;
}
.asc::after {
content: ' ↑';
}
.desc::after {
content: ' ↓';
}
#toggleButton {
background-color: #ffffff;
border: 1px solid #dddddd;
color: #555555;
padding: 10px 20px;
text-align: center;
text-decoration: none;
display: inline-block;
font-size: 14px;
margin: 4px 2px;
cursor: pointer;
border-radius: 25px;
box-shadow: 0 4px 8px 0 rgba(0,0,0,0.2);
transition-duration: 0.4s;
}
#toggleButton:hover {
box-shadow: 0 12px 16px 0 rgba(0,0,0,0.24), 0 17px 50px 0 rgba(0,0,0,0.19); /* 鼠标悬停时的阴影效果 */
}
table {
border-collapse: collapse;
width: 100%;
margin-top: 5px;
border: 1px solid #ddd;
font-size: 14px;
}
th, td {
text-align: left;
padding: 8px;
}
th {
background-color: #f2f2f2;
border-bottom: 2px solid #ddd;
}
td:hover {background-color: #ffffff;}
</style>
</body>
</html>