-
Notifications
You must be signed in to change notification settings - Fork 2
/
index.html
406 lines (351 loc) · 17.6 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
<!DOCTYPE html>
<html>
<head>
<title>IVM</title>
<style>
.hidden {
display: none;
}
</style>
<script src="https://cdn.jsdelivr.net/npm/chart.js"></script>
<meta charset="utf-8">
<meta name="description"
content="Instruction-Guided Visual Masking">
<meta name="keywords" content="MultiModal Instruction Following, Visual Grounding, Large MultiModal Models, Embodied AI">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Instruction-Guided Visual Masking</title>
<!-- <link rel="icon" href="./assets/icon.png"> -->
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" rel="stylesheet">
<link rel="stylesheet" href="./assets/css/bulma.min.css">
<link rel="stylesheet" href="./assets/css/bulma-carousel.min.css">
<link rel="stylesheet" href="./assets/css/bulma-slider.min.css">
<link rel="stylesheet" href="./assets/css/fontawesome.all.min.css">
<link rel="stylesheet" href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<script defer src="./assets/js/fontawesome.all.min.js"></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/ZhengYinan-AIR/OMIGA">
<b>OMIGA</b> <p style="font-size:18px; display: inline; margin-left: 5px;"></p>
</a>
</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="./assets/icon.png" style="width:1em;vertical-align: middle" alt="Logo"/> -->
<span class="mmmu" style="vertical-align: middle">Instruction-Guided Visual Masking</span>
</h1>
<!-- <h2 class="subtitle is-3 publication-subtitle">
Instruction-Guided Visual Masking
<br>
and Reasoning Benchmark for Expert AGI
</h2> -->
<div class="is-size-5 publication-authors">
<span class="author-block">Jinliang Zheng* <sup>†</sup> <sup style="color:#DC4437;">1</sup>,<sup style="color:#4385F5;">2</sup></span>
<span class="author-block">Jianxiong Li* <sup style="color:#DC4437;">1</sup> </span>
<span class="author-block">Sijie Cheng<sup style="color:#DC4437;">1</sup>,</span>
<span class="author-block">Yinan Zheng<sup style="color:#DC4437;">1</sup>,</span>
<span class="author-block">Jiaming Li<sup style="color:#DC4437;">1</sup>,</span>
<span class="author-block">Jihao Liu<sup style="color:#F5B400;">3</sup> <sup style="color:#4385F5;">2</sup>,</span><br>
<span class="author-block">Yu Liu<sup style="color:#4385F5;">2</sup>,</span>
<span class="author-block">Jingjing Liu<sup>✉</sup><sup style="color:#DC4437;">1</sup>,</span>
<span class="author-block">Xianyuan Zhan<sup>✉</sup> <sup style="color:#DC4437;">1</sup> <sup style="color:#f542dd;">4</sup>,</span>
</div>
<br>
<div class="is-size-5 publication-authors">
<span class="author-block"><sup style="color:#DC4437;">1</sup>AIR, Tsinghua University</span>
<span class="author-block"><sup style="color:#4385F5;">2</sup>SenseTime Research</span>
<span class="author-block"><sup style="color:#F5B400;">3</sup>CUHK MMLab</span>
<span class="author-block"><sup style="color:#f542dd;">4</sup>Shanghai AI Lab</span>
</div>
<br>
<div class="is-size-5 publication-authors">
<span class="author-block">*Equal contribution, </span><br>
<span class="author-block">†Project Lead:</span>
<span class="author-block"><a href="mailto:zhengjl23@mails.tsinghua.edu.cn">zhengjl23@mails.tsinghua.edu.cn</a></span><br>
<span class="author-block">✉Corresponding author:</span>
<span class="author-block"><a href="mailto:zhanxianyuan@air.tsinghua.edu.cn">zhanxianyuan@air.tsinghua.edu.cn</a></span>
</div>
<style>
.accepted {
text-align: center;
color: #0a283d; /* Dark blue color */
font-size: 24px;
background-color: #e8f0fe79; /* Light blue background */
border: 1px solid #B6D4FE;
border-radius: 10px;
padding: 20px;
box-shadow: 0 4px 8px rgba(0,0,0,0.1);
}
</style>
<div class="accepted">
<i class="fas fa-fire icon" style="color: red;"></i>
Exciting News!
<div>Our paper has been accepted by NeurIPS-2024</div>
</div>
<div class="accepted">
<i class="fas fa-fire icon" style="color: red;"></i>
Exciting News!
<div>Our paper has been selected as
<span style="color: red; font-weight: bold;"> outstanding paper </span>
at MFM-EAI workshop@ICML2024</div>
</div>
<div class="column has-text-centered">
<div class="publication-links">
<!-- PDF Link. -->
<span class="link-block">
<a href="https://arxiv.org/pdf/2405.19783"
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://openreview.net/forum?id=j5JvZCaDM0"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Openreview</span>
</a>
</span> -->
<!-- <span class="link-block">
<a href="https://cloud.tsinghua.edu.cn/d/0d2939f7f7234cf68f1d/"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<p style="font-size:18px">🔗</p>
</span>
<span>Dataset</span>
</a>
</span> -->
<span class="link-block">
<a href="https://github.com/2toinf/IVM"
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>
<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">
<!-- <video id="matting-video" autoplay muted controls playsinline width="100%" >
<source src="./assets/web/teaser.mp4" type="video/mp4">
</video> -->
<img src="./image/README/1716817940241.png" alt="Image" style="width: 100%;">
</div>
</div>
</section>
<section class="section">
<div class="container" style="margin-bottom: 2vh;">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h1 class="title is-2">Abstraction</h1>
<div class="content has-text-justified">
<p>
Instruction following is crucial in contemporary LLM. However, when extended
to multimodal setting, it often suffers from misalignment between specific textual
instruction and targeted local region of an image. To achieve more accurate and
nuanced multimodal instruction following, we introduce Instruction-guided Visual
Masking (IVM), a new versatile visual grounding model that is compatible with
diverse multimodal models, such as LMM and robot model. By constructing
visual masks for instruction-irrelevant regions, IVM-enhanced multimodal models
can effectively focus on task-relevant image regions to better align with complex
instructions. Specifically, we design a visual masking data generation pipeline
and create an IVM-Mix-1M dataset with 1 million image-instruction pairs. We
further introduce a new learning technique, Discriminator Weighted Supervised
Learning (DWSL) for preferential IVM training that prioritizes high-quality data
samples. Experimental results on generic multimodal tasks such as VQA and
embodied robotic control demonstrate the versatility of IVM, which as a plugand-play tool, significantly boosts the performance of diverse multimodal models,
yielding new state-of-the-art results across challenging multimodal benchmarks.
</p>
</div>
</div>
</div>
</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">
<h1 class="title is-2">Downstream Control Tasks Results</h1>
<div class="content has-text-justified">
<p>
IVM model proves valuable in
vision-language robotic manipulation tasks, where data collection is notoriously challenging and
generalization is a major concern. With the integration of IVM, our enhanced robot model
exhibits boosted performance and better generalization capabilities.
</p>
<div class="column">
<h2 style="text-align:left" class="title is-3">Results on Real Robots</h2>
<div class="content has-text-justified">
<div class="content has-text-centered">
<img src="./image/robot_score.jpg" alt="algebraic reasoning" class="center">
<p><b><i>Figure 1:</i></b> Real robot LCBC experimental results. Success rate is averaged over 10 episodes and 3 seeds.</p>
</div>
<div class="column">
<h3 style="text-align:left" class="title is-4">Red cup on red plate</h3>
<div class="content has-text-justified">
</div>
<video id="dollyzoom" autoplay controls muted loop playsinline width="48%" height="100%">
<source src="./image/video/redcup_redplate1.mp4"
type="video/mp4">
</video>
<video id="dollyzoom" autoplay controls muted loop playsinline width="48%" height="100%">
<source src="./image/video/redcup_redplate3.mp4"
type="video/mp4">
</video>
</div>
<div class="column">
<h3 style="text-align:left" class="title is-4">Duck on green plate</h3>
<div class="content has-text-justified">
</div>
<video id="dollyzoom" autoplay controls muted loop playsinline width="48%" height="100%">
<source src="./image/video/duck_greenplate1.mp4"
type="video/mp4">
</video>
<video id="dollyzoom" autoplay controls muted loop playsinline width="48%" height="100%">
<source src="./image/video/duck_greenplate2.mp4"
type="video/mp4">
</video>
</div>
<div class="column">
<h3 style="text-align:left" class="title is-4">Red cup on sivler plate</h3>
<div class="content has-text-justified">
</div>
<video id="dollyzoom" autoplay controls muted loop playsinline width="48%" height="100%">
<source src="./image/video/redcup_silverpan1.mp4"
type="video/mp4">
</video>
<video id="dollyzoom" autoplay controls muted loop playsinline width="48%" height="100%">
<source src="./image/video/redcup_silverpan5.mp4"
type="video/mp4">
</video>
</div>
<div class="column">
<h3 style="text-align:left" class="title is-4">Duck in pot</h3>
<div class="content has-text-justified">
</div>
<video id="dollyzoom" autoplay controls muted loop playsinline width="48%" height="100%">
<source src="./image/video/duck_pot1.mp4"
type="video/mp4">
</video>
<video id="dollyzoom" autoplay controls muted loop playsinline width="48%" height="100%">
<source src="./image/video/duck_pot2.mp4"
type="video/mp4">
</video>
</div>
</div>
</div>
</div>
</div>
</div>
<div class="column">
<h2 style="text-align:left" class="title is-3">Results on VQA-type benchmarks</h2>
<h3 style="text-align:left" class="title is-4">V* bench and visualization results</h3>
<div class="content has-text-justified">
<p>
We evaluate IVM-enhanced GPT4-V on
V*bench, a recently proposed challenging VQA-type benchmark characterized by images with
abundant redundancies. Results are presented in Table 1. The accuracy of the vanilla GPT4-V is
mediocre (55.0%). Our IVM model, however, can significantly improve the performance (+26.2%).
Except for the reported score, we provide more visualization results.
</p>
</div>
<img src="./image/vqa_result.png" alt="Image" style="width: 35%;">
<img src="./image/Goodcase.png" alt="Image" style="width: 60%;">
</div>
</section>
<!-- <section class="section">
<div class="container">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Feasibility-Guided Diffusion Model</h2>
<div class="content has-text-justified">
<p>
We propose a feasibility-dependent objective, i.e.,
<b>maximizing reward value within the feasible region while minimizing safety risks in the infeasible region</b>.
In FISOR, the optimal policy for the optimization problem can be derived in a special form of weighted behavior cloning.
Moreover, we propose a novel energy-guided sampling method that <b>does not require training a complicated time-dependent classifier</b> to simplify the training. <b>No more Lagrangian.</b>
</p>
<div class="content has-text-centered">
<img src="./assets/framework.jpg" alt="algebraic reasoning" class="center">
<p><b><i>Figure 2:</i></b> Feasibility-guided diffusion model with time-independent classifier-guided sampling method.</p>
</div>
</div>
</div>
</div>
</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{zheng2024instruction,
title={Instruction-Guided Visual Masking},
author={Zheng, Jinliang and Li, Jianxiong and Cheng, Sijie and Zheng, Yinan and Li, Jiaming and Liu, Jihao and Liu, Yu and Liu, Jingjing and Zhan, Xianyuan},
journal={arXiv preprint arXiv:2405.19783},
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://mmmu-benchmark.github.io/">MMMU</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>
</body>