-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
564 lines (543 loc) · 19.6 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
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
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>AIN: The Arabic Inclusive Multimodal Model</title>
<link rel="icon" type="image/png" href="images/AIN.png">
<style>
.table-container {
max-width: 60%; /* Matches the fig-container width */
margin: 30px auto;
overflow-x: auto;
}
.performance-table {
width: 100%;
border-collapse: collapse;
margin: 20px 0;
font-size: 0.9em;
}
.performance-table th,
.performance-table td {
border: 1px solid #ddd;
padding: 8px;
text-align: center;
}
.performance-table th {
background-color: #e0e0e0;
font-weight: bold;
}
.performance-table tr:nth-child(even) {
background-color: #f8f8f8;
}
.performance-table td.total-column {
background-color: #d0d0d0;
}
.performance-table tr:hover {
background-color: #f5f5f5;
}
.table-caption {
width: 100%;
text-align: center;
margin-top: 10px;
font-style: italic;
color: #666;
font-size: 0.9em;
}
.badge-button {
display: inline-flex;
align-items: center;
gap: 0.5rem;
padding: 0.5rem 1rem;
background-color: #7a7a7a;
color: white;
border-radius: 0.5rem;
text-decoration: none;
margin-right: 1rem;
transition: background-color 0.2s;
}
.badge-button:hover {
background-color: #3d9399;
}
.badge-button img {
width: 1.5rem;
height: 1.5rem;
}
body {
position: relative;
}
.demo-links {
text-align: center;
margin: 10px 0;
background-color: #e6e6e6;
padding: 20px;
border-radius: 10px;
}
.demo-links a {
color: #333;
margin: 0 20px;
text-decoration: none;
display: inline-flex;
align-items: center;
}
.demo-links img {
width: 40px;
height: 40px;
margin-right: 8px;
}
body {
font-family: Arial, sans-serif;
margin: 0;
padding: 20px;
line-height: 1.6;
color: #333;
background-color: white;
}
.container {
max-width: 1200px;
margin: 0 auto;
padding: 20px;
background-color: white;
box-shadow: white;
border-radius: 8px;
}
.container1 {
max-width: 1200px;
margin: 0 auto;
padding: 10px;
background-color: linear-gradient(to right, #b8fbe9, #c2f9fe);
border-radius: 10px;
}
.header {
text-align: center;
margin-bottom: 30px;
}
.logo {
width:22%;
display: block;
margin: -5px auto;
}
.container {
text-align: center;
}
h1, h2 {
text-align: center;
margin: 30px 0;
}
.fig-container img {
max-width: 65%; /* Reduced from 80% */
height: auto;
margin: 0 auto;
display: block;
}
.fig-container-side {
text-align: center;
margin: 0px 0;
}
.fig-container-side img:first-child {
width: 50%;
margin: 0 2%;
display: inline-block;
}
.fig-container-side img:last-child {
width: 40%;
margin: 0 2%;
display: inline-block;
}
p {
text-align: center;
max-width: 800px;
margin: 20px auto;
}
ul {
display: inline-block;
text-align: left;
margin: 0 auto;
}
.features {
text-align: center;
}
.features ul {
padding-left: 20px;
}
.badges {
text-align: center;
}
.publication-authors {
text-align: center;
max-width: 800px;
margin: 20px auto;
}
.citation {
text-align: left;
max-width: 600px;
margin: 20px auto;
padding: 20px;
}
.title {
font-size: 6.5em;
margin: 0px 0;
background: linear-gradient(to right, #333333, #666666);
-webkit-background-clip: text;
-webkit-text-fill-color: transparent;
font-weight: bold;
font-family:Geneva, Tahoma, sans-serif
}
.title2 {
font-size:2.5em;
margin: 0px 0;
background: linear-gradient(to right, #585858, #666666);
-webkit-background-clip: text;
-webkit-text-fill-color: transparent;
font-family:Geneva, Tahoma, sans-serif
}
.authors {
font-size: 18px;
margin: 20px 0;
}
.badges {
margin: 20px 0;
}
.badges img {
margin: 0 5px;
max-width: 120%;
}
.fig-container {
text-align: center;
margin: 30px 0;
}
.fig-container img {
max-width: 60%;
height: auto;
}
.fig-container-s {
text-align: center;
margin: 30px 0;
}
.fig-container-s img {
max-width: 40%;
}
.caption {
font-style: italic;
color: #666;
margin-top: 10px;
font-size: 0.9em;
}
table {
width: 100%;
border-collapse: collapse;
margin: 20px 0;
}
th, td {
border: 1px solid #ddd;
padding: 8px;
text-align: center;
}
th {
background-color: #e0e0e0;
}
.logos {
display: flex;
justify-content: center;
align-items: center;
gap: 20px;
margin-top: 40px;
}
.logos img {
height: 50px;
}
a {
color: #007bff;
text-decoration: none;
}
a:hover {
text-decoration: underline;
}
.features {
margin: 30px 0;
}
.section {
margin: 20px 0;
}
.citation {
background-color: #f8f9fa;
padding: 20px;
border-radius: 4px;
font-family: monospace;
}
.abstract-container {
max-width: 900px;
margin: 20px auto;
padding: 20px 20px;
background-color:white;
}
.abstract-container h2 {
text-align: center;
margin-bottom: 20px;
color: #333;
}
.abstract-content {
text-align: justify;
line-height: 1.6;
margin: 0;
color: #2c2c2c;
}
.abstract-content p {
margin-bottom: 15px;
}
.abstract-content p:last-child {
margin-bottom: 0;
}
</style>
</head>
<body>
<img src="images/AIN_logo_git_5.png" alt="AIN Logo" class="logo">
<!-- <img src="images/AIN.png" alt="AIN Logo" class="logo"> <h1 class="title">AIN</h1>-->
<h1 class="title2">The Arabic INclusive Multimodal Model</h1>
<!-- Authors -->
<div class="authors">
<p>
Ahmed Heakl<sup style="color:#3399FF;">1*</sup>,
Sara Ghaboura<sup style="color:#3399FF;">1*</sup>,
Omkar Thawakar<sup style="color:#3399FF;">1</sup>, <br>
Fahad S. Khan<sup style="color:#3399FF;">1, </sup><sup style="color:#FF66B3;">2</sup>,
Hisham Cholakkal<sup style="color:#3399FF;">1</sup>,
Rao M. Anwer<sup style="color:#3399FF;">1, </sup><sup style="color:#4CB5AE;">3</sup>,
Salman Khan<sup style="color:#3399FF;">1, </sup><sup style="color:#FFA280;">4</sup>
</p>
<p>
<sup style="color:#3399FF;">1</sup>Mohamed bin Zayed University of AI,
<sup style="color:#4CB5AE;">3</sup>Aalto University,
<sup style="color:#FF66B3;">2</sup>Linköping University,
<sup style="color:#FFA280;">4</sup>Australian National University
</p>
</div>
<div class="badges">
<a href="https://arxiv.org/abs/2502.00094" class="badge-button">
<img src="images/pdf.png" alt="Paper"> Paper
</a>
<a href="https://arxiv.org/abs/2502.00094" class="badge-button">
<img src="images/arxiv.png" alt="AIN"> arxiv
</a>
<a href="https://github.com/mbzuai-oryx/AIN" class="badge-button">
<img src="images/github.png" alt="Code"> Github
</a>
</div>
<div class="container1">
<div class="demo-links">
<a href="https://huggingface.co/spaces/ahmedheakl/AIN-Arabic-VLM" target="_blank">
<img src="https://cdn-icons-png.flaticon.com/512/1698/1698535.png" alt="Demo">
AIN Chatbot
</a>
<a href="https://t.me/arabicvlm_bot" target="_blank">
<img src="https://cdn-icons-png.flaticon.com/512/2111/2111646.png" alt="Telegram">
AIN Telegram
</a>
<a href="https://wa.me/46738645096" target="_blank">
<img src="https://cdn-icons-png.flaticon.com/512/3670/3670051.png" alt="WhatsApp">
AIN WhatsApp
</a>
</div>
</div>
<!-- AIN Can See Figure -->
<div class="fig-container">
<img src="images/ain_can_see.png" alt="AIN Can See">
<p class="caption">
AIN: A versatile LMM excelling in visual and contextual understanding across diverse domains, including VQA on complex topics, OCR for various fonts and handwriting, cultural insights (traditions, food, places), agricultural tasks (crop identification, fruit classification, disease detection), remote sensing (multi-scale objects), medical imaging (various modalities), and video analysis (animation, human activities)
</p>
</div>
<!-- Abstract -->
<div class="abstract-container">
<h2>Abstract</h2>
<div class="abstract-content">
<p>
Amid the swift progress of large language models (LLMs) and their evolution into large multimodal models (LMMs), significant strides have been made in high-resource languages such as English and Chinese. While Arabic LLMs have seen notable progress, Arabic LMMs remain largely unexplored, often narrowly focusing on a few specific aspects of the language and visual understanding. To bridge this gap, we introduce <b><em>AIN - the Arabic Inclusive Multimodal Model-</em></b> designed to excel across diverse domains.
AIN is an English-Arabic <b>bilingual LMM</b> designed to excel in English and Arabic, leveraging carefully constructed <b>3.6 million</b> high-quality Arabic-English multimodal data samples. AIN demonstrates state-of-the-art Arabic performance, while also possessing strong English-language visual capabilities.
</p>
</div>
</div>
<div align="center">
<video width="620" height="360" controls>
<source src="images/AIN_teaser_2.mp4" type="video/mp4">
</video>
</div>
<!-- Latest Updates -->
<div class="section">
<h2>📢 Latest Updates</h2>
<p>🔥 <b>Jan 2025</b> AIN-7B model, the first Arabic Inclusive LMM, is released 🤗<br>
🚀 Model weights will be released soon.</p>
</div>
<!-- Key Features -->
<div class="features">
<h1>🌟 Key Features</h1>
<ul>
<li>First Arabic-centric inclusive Large Multimodal Model (LMM) trained on 3.6M samples</li>
<li>Includes 35% authentic Arabic data within its Arabic data subset</li>
<li>Achieves superior performance compared to both open- and closed-source models</li>
<li>Demonstrates robust bilingual capabilities (Arabic/English)</li>
<li>Exhibits advanced cultural understanding and domain expertise</li>
</ul>
</div>
<!-- Performance Analysis Figures -->
<div class="fig-container-s">
<img src="images/radar_chart.png" alt="Performance Analysis">
<p class="caption"><strong> Figure 1:</strong> The chart showcases a comprehensive performance analysis of AIN-7B across CAMEL-Bench domains, comparing it with prominent closed-source models and open-source counterparts.<em><b>OCR:</b> "OCR & Document Understanding", <b>Video:</b> "General Video & Multi-Image Understanding", <b>RS:</b> "Remote Sensing Understanding", <b>CDT:</b> "Chart, Diagram & Table Understanding", <b>Agro.:</b> "Agricultural Image Understanding", <b>Cultural:</b> "Cultural-Specific Understanding", <b>Medical:</b> "Medical Image Understanding".</em></p>
</div>
<div class="fig-container">
<img src="images/intro_bar.png" alt="Comparative Performance">
<p class="caption"> <strong>Figure 2:</strong> Comparative performance of AIN-7B against other models across key domains, including OCR & Document Understanding, Remote Sensing, Agricultural Understanding, and overall performance across all domains.</p>
</div>
<div class="table-container">
<table class="performance-table">
<thread>
<tr style="background-color: #e0e0e0;">
<th>Models</th>
<th>VQA</th>
<th>OCR</th>
<th>Video</th>
<th>RS</th>
<th>CDT</th>
<th>Agro.</th>
<th>Cult.</th>
<th>Med.</th>
<th style="background-color: #d0d0d0;">Total</th>
</tr>
</thead>
<tbody>
<tr>
<td>GPT-4o</td>
<td>🥈55.15</td>
<td>🥈54.98</td>
<td>🥇69.65</td>
<td>🥈27.36</td>
<td>🥈62.35</td>
<td>🥈80.75</td>
<td>🥇80.86</td>
<td>🥇49.91</td>
<td style="background-color: #d0d0d0;">🥈60.13</td>
</tr>
<tr>
<td>GPT-4o-mini</td>
<td>48.83</td>
<td>39.38</td>
<td>🥈 66.28</td>
<td>16.93</td>
<td>56.37</td>
<td>78.80</td>
<td>65.92</td>
<td>🥈 47.37</td>
<td style="background-color: #d0d0d0;">52.49</td>
</tr>
<tr>
<td>Gemini-1.5-Pro</td>
<td>46.68</td>
<td>28.68</td>
<td>42.95</td>
<td>17.07</td>
<td>47.06</td>
<td>72.14</td>
<td>56.24</td>
<td>33.78</td>
<td style="background-color: #d0d0d0;">52.38</td>
</tr>
<tr>
<td>Gemini-1.5-flash</td>
<td>45.59</td>
<td>27.58</td>
<td>53.31</td>
<td>14.95</td>
<td>48.26</td>
<td>76.07</td>
<td>46.54</td>
<td>42.87</td>
<td style="background-color: #d0d0d0;">44.40</td>
</tr>
<tr>
<td>InternVL-8B </td>
<td>30.41 </td>
<td>15.91 </td>
<td>51.42 </td>
<td>5.36 </td>
<td>30.27 </td>
<td>44.47 </td>
<td>20.88 </td>
<td>29.48 </td>
<td style="background-color: #d0d0d0;">28.52 </td>
</tr>
<tr>
<td>InternVL2.5-1B </td>
<td>27.22 </td>
<td>19.45 </td>
<td>38.20 </td>
<td>3.39 </td>
<td>30.75 </td>
<td>39.53 </td>
<td>35.68 </td>
<td>21.27 </td>
<td style="background-color: #d0d0d0;">26.94 </td>
</tr>
<tr>
<td>Qwen-VL-2B </td>
<td>41.02 </td>
<td>22.93 </td>
<td>38.90 </td>
<td>12.56 </td>
<td>27.83 </td>
<td>52.02 </td>
<td>34.28 </td>
<td>29.12 </td>
<td style="background-color: #d0d0d0;">32.33 </td>
</tr>
<tr>
<td>AIN-7B <em>(ours)</em> </td>
<td>🥇56.78 </td>
<td>🥇72.35 </td>
<td>64.09 </td>
<td>🥇45.92 </td>
<td>🥇64.10 </td>
<td>🥇85.05 </td>
<td>🥈78.09 </td>
<td>43.77 </td>
<td style="background-color: #d0d0d0;">🏆63.77 </td>
</tr>
</tbody>
</table>
<p class="table-caption">
<strong>Table 1.</strong> Performance comparison of AIN and different closed- and open-source LMMs across CAMEL-Bench domains.<em>Best performance is marked with 🥇; second-best is 🥈.</em>
</p>
</div>
<!-- Qualitative Analysis -->
<div class="fig-container-s fig-container-side">
<!-- <img src="images/contre.png" alt="Qualitative Comparison"> -->
<img src="images/qualitative.png" alt="Qualitative Examples">
<p class="caption"><strong>Figure 3:</strong> Qualitative examples showcasing AIN-7B’s capabilities across various domains, including general VQA, OCR & Document Understanding, Remote Sensing, Medical Imaging, Agricultural Understanding, and Cultural-Specific tasks.</p>
</div>
<!-- Verification Pipeline -->
<div class="section">
<h2>🧐 Data Verification and Toxicity Filtering</h2>
<div class="fig-container">
<img src="images/verify_pipeline.png" alt="Verification Pipeline">
<p class="caption"><strong>Figure 4:</strong> Data verification and filtering pipeline for textual and visual data, ensuring high-quality training data through semantic similarity checks, translation quality evaluations, and toxicity screening for safety compliance.</p>
</div>
</div>
<!-- Citation -->
<div class="section">
<h2>📚 Citation</h2>
<div class="citation">
@misc{heakl2025ainarabicinclusivelarge, <br>
title={AIN: The Arabic INclusive Large Multimodal Model}, <br>
author={Ahmed Heakl and Sara Ghaboura and Omkar Thawkar and Fahad Shahbaz Khan and Hisham Cholakkal and Rao Muhammad Anwer and Salman Khan},<br>
year={2025},<br>
url={https://arxiv.org/abs/2502.00094},<br>
}
</div>
</div>
<!-- Logos -->
<div class="logos">
<img src="images/IVAL_logo.png" alt="IVAL Logo">
<img src="images/Oryx_logo.png" alt="Oryx Logo">
<img src="images/MBZUAI_logo.png" alt="MBZUAI Logo">
</div>
</div>
</body>
</html>