-
Notifications
You must be signed in to change notification settings - Fork 3
/
index.html
558 lines (503 loc) · 25.8 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
<!DOCTYPE html>
<head>
<!-- Global site tag (gtag.js) - Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=UA-178132094-1"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag() {
dataLayer.push(arguments);
}
gtag("js", new Date());
gtag("config", "UA-178132094-1");
</script>
<meta charset="UTF-8" />
<!-- <meta name="viewport" content="width=device-width, initial-scale=1" />-->
<meta name="viewport" content="width=1024" />
<title>RobustBench: Adversarial robustness benchmark</title>
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script src="https://maxcdn.bootstrapcdn.com/bootstrap/4.3.1/js/bootstrap.min.js"></script>
<link rel="stylesheet" href="https://stackpath.bootstrapcdn.com/bootstrap/4.3.1/css/bootstrap.min.css" integrity="sha384-ggOyR0iXCbMQv3Xipma34MD+dH/1fQ784/j6cY/iJTQUOhcWr7x9JvoRxT2MZw1T" crossorigin="anonymous">
<script src="https://polyfill.io/v3/polyfill.min.js?features=es6"></script>
<script type="text/javascript" async
src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.7/MathJax.js?config=TeX-MML-AM_CHTML">
</script>
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/foundation/6.4.3/css/foundation.min.css" />
<link rel="stylesheet" href="https://cdn.rawgit.com/jpswalsh/academicons/master/css/academicons.min.css" />
<script src="https://kit.fontawesome.com/b939870cfb.js" crossorigin="anonymous"></script>
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/foundation/6.4.3/css/foundation.min.css">
<link rel="stylesheet" href="https://cdn.datatables.net/1.10.24/css/dataTables.foundation.min.css">
<script type="text/javascript" src="https://code.jquery.com/jquery-3.5.1.js"></script>
<script type="text/javascript" src="https://cdn.datatables.net/1.10.24/js/jquery.dataTables.min.js"></script>
<script type="text/javascript" src="https://cdn.datatables.net/1.10.24/js/dataTables.foundation.min.js"></script>
<link rel="stylesheet" href="./css/main.css" />
</head>
<body>
<nav class="navbar navbar-expand-md">
<div class="container">
<a class="navbar-brand" href="./index.html"
>RobustBench</a>
<button
class="navbar-toggler navbar-light"
type="button"
data-toggle="collapse"
data-target="#main-navigation"
>
<span class="navbar-toggler-icon"></span>
</button>
<div class="collapse navbar-collapse" id="main-navigation">
<ul class="navbar-nav">
<li class="nav-item">
<a class="nav-link" href="#leaderboard">Leaderboards</a>
</li>
<li>
<a class="nav-link" href="https://arxiv.org/abs/2010.09670">Paper</a>
</li>
<li>
<a class="nav-link" href="#faq"
>FAQ</a
>
</li>
<li>
<a class="nav-link" href="#contribute">Contribute</a>
</li>
<li>
<a class="nav-link text-nowrap" href="https://github.com/RobustBench/robustbench"
>Model Zoo 🚀</a
>
</li>
</ul>
</div>
</div>
</nav>
<!-- <hr class="toprule" /> -->
<header>
<div class="header-block container">
<div class="logo"><img src="./images/logo.png" alt="logo" /></div>
<div class="title">RobustBench</div>
<div class="description">
A standardized benchmark for adversarial robustness
</div>
</div>
</header>
<!-- <hr class="toprule" /> -->
<div class="container">
<section id="introduction">
<div class="overview">
<p class="doublealign">
The goal of <strong>RobustBench</strong> is to systematically track
the <em>real</em> progress in adversarial robustness. There are
already
<a href="https://nicholas.carlini.com/writing/2019/all-adversarial-example-papers.html">more than 3'000
papers</a>
on this topic, but it is still unclear which approaches really work
and which only lead to
<a href="https://arxiv.org/abs/1802.00420">overestimated robustness</a>. We start from benchmarking common
corruptions,
\(\ell_\infty\)- and
\(\ell_2\)-robustness since these are the most studied settings in the
literature. We use
<a href="https://github.com/fra31/auto-attack">AutoAttack</a>, an
ensemble of white-box and black-box attacks, to standardize the
evaluation (for details see <a href="https://arxiv.org/abs/2010.09670">our paper</a>) of the \(\ell_p\)
robustness and
CIFAR-10-C for the evaluation of robustness to common corruptions. Additionally,
we open source the
<a href="https://github.com/RobustBench/robustbench">RobustBench library</a>
that contains models used for the leaderboard to facilitate their
usage for downstream applications. <br><br>
To prevent potential overadaptation of new defenses to AutoAttack, we also welcome external evaluations based
on <em>adaptive attacks</em>, especially where AutoAttack <a href="https://github.com/fra31/auto-attack/blob/master/flags_doc.md">flags</a>
a potential overestimation of robustness. For each model, we are interested in the best known robust accuracy
and see AutoAttack and adaptive attacks as complementary.
<br><br>
<strong> News:</strong>
<ul>
<li> <strong>May 2022:</strong>
We have extended the common corruptions leaderboard on ImageNet with <a href="https://3dcommoncorruptions.epfl.ch">3D Common Corruptions</a> (ImageNet-3DCC). ImageNet-3DCC evaluation is interesting since (1) it includes more realistic corruptions and (2) it can be used to assess generalization of the existing models which may have overfitted to ImageNet-C. For a quickstart, click <a href="https://github.com/RobustBench/robustbench#new-evaluating-robustness-of-imagenet-models-against-3d-common-corruptions-imagenet-3dcc">here</a>. See the new leaderboard with ImageNet-C and ImageNet-3DCC <a href="https://robustbench.github.io/#div_imagenet_corruptions_heading">here</a> (also mCE metrics can be found <a href="https://github.com/RobustBench/robustbench#corruptions-imagenet-c--imagenet-3dcc">here</a>).
</li>
<li> <strong>May 2022:</strong>
We fixed the preprocessing issue for ImageNet corruption evaluations: previously we used resize to 256x256 and central crop to 224x224 which wasn't necessary since the ImageNet-C images are already 224x224. Note that this changed the ranking between the top-1 and top-2 entries.
</li>
</ul>
</p>
<div class="flexbox-container features">
<div class="element">
<div class="icon">
<img src="https://img.icons8.com/wired/100/000000/leaderboard.png" />
</div>
<p>
Up-to-date leaderboard based <br />
on 120+ models
</p>
</div>
<div class="element">
<div class="icon">
<img src="https://img.icons8.com/ios-glyphs/80/000000/user-credentials.png" />
</div>
<p>
Unified access to 80+ state-of-the-art <br />robust models via
Model Zoo
</p>
</div>
</div>
</div>
<div class="details">
<div class="box usage">
<p>Model Zoo</p>
<div class="divider">
<hr />
</div>
Check out the
<a href="https://github.com/RobustBench/robustbench#model-zoo">available models</a>
and our
<a href="https://github.com/RobustBench/robustbench#notebooks">Colab tutorials</a>.
<div class="codeblock">
<div class="vspace10"></div>
<!--
# !pip install git+https://github.com/RobustBench/[email protected]
from robustbench.utils import load_model
# Load a model from the model zoo
model = load_model(model_name='Rebuffi2021Fixing_70_16_cutmix_extra',
dataset='cifar10',
threat_model='Linf')
# Evaluate the Linf robustness of the model using AutoAttack
from robustbench.eval import benchmark
clean_acc, robust_acc = benchmark(model,
dataset='cifar10',
threat_model='Linf')
-->
<!-- HTML generated using hilite.me -->
<div style="background: #ffffff; overflow:auto;width:auto;padding:.2em .6em;">
<pre style="margin: 0; line-height: 125%"><span style="color: #888888"># !pip install git+https://github.com/RobustBench/[email protected]</span>
<span style="color: #008800; font-weight: bold">from</span> <span style="color: #0e84b5; font-weight: bold">robustbench.utils</span> <span style="color: #008800; font-weight: bold">import</span> load_model
<span style="color: #888888"># Load a model from the model zoo</span>
model <span style="color: #333333">=</span> load_model(model_name<span style="color: #333333">=</span><span style="background-color: #fff0f0">'Rebuffi2021Fixing_70_16_cutmix_extra'</span>,
dataset<span style="color: #333333">=</span><span style="background-color: #fff0f0">'cifar10'</span>,
threat_model<span style="color: #333333">=</span><span style="background-color: #fff0f0">'Linf'</span>)
<span style="color: #888888"># Evaluate the Linf robustness of the model using AutoAttack</span>
<span style="color: #008800; font-weight: bold">from</span> <span style="color: #0e84b5; font-weight: bold">robustbench.eval</span> <span style="color: #008800; font-weight: bold">import</span> benchmark
clean_acc, robust_acc <span style="color: #333333">=</span> benchmark(model,
dataset<span style="color: #333333">=</span><span style="background-color: #fff0f0">'cifar10'</span>,
threat_model<span style="color: #333333">=</span><span style="background-color: #fff0f0">'Linf'</span>)
</pre>
</div>
<!-- HTML generated using hilite.me -->
<div style="
background: #ffffff;
overflow: auto;
width: auto;
border: solid gray;
border-width: 0em 0em 0em 0em;
padding: 0.2em 0.6em;
">
</pre>
</div>
</div>
</div>
<div class="box images">
<p>Analysis</p>
<div class="divider">
<hr />
</div>
Check out <a href="https://arxiv.org/abs/2010.09670">our paper</a> with a detailed analysis.
<div>
<!-- <div class="scroller analysis-images">-->
<img class="analysis" src="./images/aa_robustness_vs_venues_Linf.png" alt="robustness_vs_venues" />
<!-- <img-->
<!-- src="./images/aa_robustness_vs_standard_Linf.png"-->
<!-- alt="robustness_vs_clean"-->
<!-- />-->
</div>
</div>
</div>
<div class="vspace10"></div>
</section>
<div id="leaderboard" class="container button-list">
<div class="heading">
<u>
Available Leaderboards
</u>
</div>
<a class="btn btn-secondary" href="#div_cifar10_Linf_heading">CIFAR-10 (\( \ell_\infty\))</a>
<a class="btn btn-secondary" href="#div_cifar10_L2_heading">CIFAR-10 (\( \ell_2\))</a>
<a class="btn btn-secondary" href="#div_cifar10_corruptions_heading">CIFAR-10 (Corruptions)</a>
<a class="btn btn-secondary" href="#div_cifar100_Linf_heading">CIFAR-100 (\( \ell_\infty\))</a>
<a class="btn btn-secondary" href="#div_cifar100_corruptions_heading">CIFAR-100 (Corruptions)</a>
<a class="btn btn-secondary" href="#div_imagenet_Linf_heading">ImageNet (\( \ell_\infty\))</a>
<a class="btn btn-secondary" href="#div_imagenet_corruptions_heading">ImageNet (Corruptions: IN-C, IN-3DCC)</a>
</div>
<section class="container" id="div_cifar10_Linf_heading">
<div class="heading">
<p>
<!-- <div style="opacity:0;">1<br>1<br>1<br>1</div> <!– Needed to introduce a margin for the navigation bar between leaderboards –>-->
Leaderboard:
<span class="heading-math">CIFAR-10, \( \ell_\infty = 8/255 \)</span>,
untargeted attack
</p>
</div>
<div id="div_cifar10_Linf"></div>
</section>
<section>
<div class="heading" id="div_cifar10_L2_heading">
<p>
<div style="opacity:0;">1<br>1<br>1<br>1</div> <!-- Needed to introduce a margin for the navigation bar between leaderboards -->
Leaderboard:
<span class="heading-math">CIFAR-10, \( \ell_2 = 0.5 \)</span>,
untargeted attack
</p>
</div>
<div id="div_cifar10_L2"></div>
</section>
<section>
<div class="heading" id="div_cifar10_corruptions_heading">
<p>
<div style="opacity:0;">1<br>1<br>1<br>1</div> <!-- Needed to introduce a margin for the navigation bar between leaderboards -->
Leaderboard:
<span class="heading-math">CIFAR-10, Common Corruptions</span>,
CIFAR-10-C
</p>
</div>
<div id="div_cifar10_corruptions"></div>
</section>
<section>
<div class="heading" id="div_cifar100_Linf_heading">
<p>
<div style="opacity:0;">1<br>1<br>1<br>1</div> <!-- Needed to introduce a margin for the navigation bar between leaderboards -->
Leaderboard:
<span class="heading-math">CIFAR-100, \( \ell_\infty = 8/255 \)</span>,
untargeted attack
</p>
</div>
<div id="div_cifar100_Linf"></div>
</section>
<section>
<div class="heading" id="div_cifar100_corruptions_heading">
<p>
<div style="opacity:0;">1<br>1<br>1<br>1</div> <!-- Needed to introduce a margin for the navigation bar between leaderboards -->
Leaderboard:
<span class="heading-math">CIFAR-100, Common Corruptions</span>,
CIFAR-100-C
</p>
</div>
<div id="div_cifar100_corruptions"></div>
</section>
<section>
<div class="heading" id="div_imagenet_Linf_heading">
<p>
<div style="opacity:0;">1<br>1<br>1<br>1</div> <!-- Needed to introduce a margin for the navigation bar between leaderboards -->
Leaderboard:
<span class="heading-math">ImageNet, \( \ell_\infty = 4/255 \)</span>,
untargeted attack
</p>
</div>
<div id="div_imagenet_Linf"></div>
</section>
<section>
<div class="heading" id="div_imagenet_corruptions_heading">
<p>
<div style="opacity:0;">1<br>1<br>1<br>1</div> <!-- Needed to introduce a margin for the navigation bar between leaderboards -->
Leaderboard:
<span class="heading-math">ImageNet, Common Corruptions (ImageNet-C, ImageNet-3DCC)</span>
</p>
</div>
<div id="div_imagenet_corruptions"></div>
</section>
<div class="vspace30"></div>
<section id="faq">
<div class="heading">
<p>FAQ</p>
</div>
<p class="qa-box">
<span class="question">➤ How does the RobustBench leaderboard differ from the
<a href="https://github.com/fra31/auto-attack">AutoAttack leaderboard</a>? 🤔
</span>
<br />
<span class="answer"> The <a href="https://github.com/fra31/auto-attack">AutoAttack leaderboard</a> was
the starting point of RobustBench. Now only the RobustBench leaderboard is actively maintained.
</span>
</p>
<p class="qa-box">
<span class="question">➤ How does the RobustBench leaderboard differ from
<a href="https://www.robust-ml.org/">robust-ml.org</a>? 🤔
</span>
<br />
<span class="answer"><a href="https://www.robust-ml.org/">robust-ml.org</a> focuses on
<em>adaptive</em> evaluations, but we provide a
<strong>standardized benchmark</strong>. Adaptive evaluations have been very useful (e.g., see
<a href="https://arxiv.org/abs/2002.08347">Tramer et al., 2020</a>),
but they are also very time-consuming and cannot be standardized by definition. Instead, we argue
that one can estimate robustness accurately mostly <em>without</em> adaptive
attacks but for this one has to introduce some restrictions on the
considered models (see <a href="https://arxiv.org/abs/2010.09670">our paper</a> for more details).
However, we do welcome adaptive evaluations and we are always interested in showing the best known
robust accuracy.
</span>
</p>
<p class="qa-box">
<span class="question">➤ How is it related to libraries like
<a href="https://github.com/bethgelab/foolbox">foolbox</a> /
<a href="https://github.com/tensorflow/cleverhans">cleverhans</a> /
<a href="https://github.com/BorealisAI/advertorch">advertorch</a>? 🤔
</span>
<br />
<span class="answer">These libraries provide implementations of different
<em>attacks</em>. Besides the standardized benchmark,
<strong>RobustBench</strong> additionally provides a repository of the
most robust models. So you can start using the robust models in one
line of code (see the tutorial
<a href="https://github.com/RobustBench/robustbench#model-zoo-quick-tour">here</a>).</span>
</p>
<p class="qa-box">
<span class="question">➤ Why is Lp-robustness still interesting? 🤔
</span>
<br />
<span class="answer">There are numerous interesting applications of Lp-robustness that
span transfer learning (<a href="https://arxiv.org/abs/2007.08489">Salman et al. (2020)</a>,
<a href="https://arxiv.org/abs/2007.05869">Utrera et al. (2020)</a>),
interpretability (<a href="https://arxiv.org/abs/1805.12152">Tsipras et al. (2018)</a>, <a
href="https://arxiv.org/abs/1910.08640">Kaur et al. (2019)</a>,
<a href="https://arxiv.org/abs/1906.00945">Engstrom et al. (2019)</a>), security (<a
href="https://arxiv.org/abs/1811.03194">Tramèr et al. (2018)</a>,
<a href="https://arxiv.org/abs/1906.07153">Saadatpanah et al. (2019)</a>), generalization (<a
href="https://arxiv.org/abs/1911.09665">Xie et al. (2019)</a>, <a
href="https://arxiv.org/abs/1909.11764">Zhu et al. (2019)</a>,
<a href="https://arxiv.org/abs/2004.10934">Bochkovskiy et al. (2020)</a>), robustness to unseen perturbations
(<a href="https://arxiv.org/abs/1911.09665">Xie et al. (2019)</a>, <a
href="https://arxiv.org/abs/1905.01034">Kang et al. (2019)</a>),
stabilization of GAN training (<a href="https://arxiv.org/abs/2008.03364">Zhong et al. (2020)</a>).</span>
</p>
<p class="qa-box">
<span class="question">➤ What about verified adversarial robustness? 🤔
</span>
<br />
<span class="answer">We mostly focus on defenses which improve
<em>empirical</em> robustness, given the lack of clarity regarding
which approaches really improve robustness and which only make some
particular attacks unsuccessful.
However, we do not restrict submissions of verifiably robust models (e.g., we have
<a href="https://arxiv.org/abs/1906.06316">Zhang et al. (2019)</a> in our CIFAR-10 Linf leaderboard).
For methods targeting verified robustness, we encourage the readers to check out
<a href="https://arxiv.org/abs/1902.08722">Salman et al. (2019)</a>
and <a href="https://arxiv.org/abs/2009.04131">Li et al. (2020)</a>.
</span>
</p>
<p class="qa-box">
<span class="question">➤ What if I have a better attack than the one used in this
benchmark? 🤔
</span>
<br />
<span class="answer">We will be happy to add a better attack or any adaptive evaluation
that would complement our default standardized attacks!</span>
</p>
</section>
<div class="vspace50"></div>
<section id="citation">
<div class="heading">
<p>Citation</p>
</div>
Consider citing our whitepaper if you want to reference our leaderboard or if you are using the models from the
Model Zoo:
<!-- @article{croce2020robustbench,-->
<!-- title={RobustBench: a standardized adversarial robustness benchmark},-->
<!-- author={Croce, Francesco and Andriushchenko, Maksym and Sehwag, Vikash and Flammarion, Nicolas and Chiang, Mung and Mittal, Prateek and Matthias Hein},-->
<!-- journal={arXiv preprint arXiv:2010.09670},-->
<!-- year={2020}-->
<!-- }-->
<!-- HTML generated using hilite.me -->
<div
style="background: #ffffff; overflow:auto;width:auto;border:solid gray;border-width:.0em .0em .0em .0em;padding:.2em .6em;">
<pre style="margin: 0; line-height: 125%"><span style="color: #555555; font-weight: bold">@article</span>{croce2020robustbench,
title<span style="color: #333333">=</span>{RobustBench: a standardized adversarial robustness benchmark},
author<span style="color: #333333">=</span>{Croce, Francesco <span style="color: #000000; font-weight: bold">and</span> Andriushchenko, Maksym <span style="color: #000000; font-weight: bold">and</span> Sehwag, Vikash <span style="color: #000000; font-weight: bold">and</span> Debenedetti, Edoardo <span style="color: #000000; font-weight: bold">and</span> Flammarion, Nicolas
<span style="color: #000000; font-weight: bold">and</span> Chiang, Mung <span style="color: #000000; font-weight: bold">and</span> Mittal, Prateek <span style="color: #000000; font-weight: bold">and</span> Matthias Hein},
journal<span style="color: #333333">=</span>{arXiv preprint arXiv:2010.09670},
year<span style="color: #333333">=</span>{2020}
}</pre>
</div>
</section>
<div class="vspace50"></div>
<section id="contribute">
<div class="details">
<div class="box2">
<p>Contribute to RobustBench!</p>
<div class="divider">
<hr />
</div>
We welcome any contribution in terms of both new robust models and
evaluations. Please check
<a href="https://github.com/RobustBench/robustbench#how-to-contribute">here</a>
for more details.
<br />
<br />
Feel free to contact us at
<a href="mailto:[email protected]">[email protected]</a>
</div>
<div class="box2">
<p>Maintainers</p>
<div class="divider">
<hr />
</div>
<ul>
<li>
<a href="https://twitter.com/fra__31" target="_blank">Francesco Croce
</a>
<a href="https://twitter.com/fra__31"><i class="fas fa-globe"></i></a>
<a href="https://github.com/fra31"><i class="fab fa-github"></i></a>
<a href="https://scholar.google.com/citations?user=laq9cq0AAAAJ"><i class="ai ai-google-scholar"></i></a>
</li>
<li>
<a href="https://people.epfl.ch/maksym.andriushchenko" target="_blank">Maksym Andriushchenko</a>
<a href="https://people.epfl.ch/maksym.andriushchenko"><i class="fas fa-globe"></i></a>
<a href="https://github.com/max-andr"><i class="fab fa-github"></i></a>
<a href="https://scholar.google.com/citations?user=ZNtuJYoAAAAJ"><i class="ai ai-google-scholar"></i></a>
</li>
<li>
<a href="https://vsehwag.github.io/" target="_blank">Vikash Sehwag</a>
<a href="https://vsehwag.github.io/"><i class="fas fa-globe"></i></a>
<a href="https://github.com/VSehwag"><i class="fab fa-github"></i></a>
<a href="https://scholar.google.com/citations?user=JAkeEG8AAAAJ"><i class="ai ai-google-scholar"></i></a>
</li>
<li>
<a href="https://edoardo.science" target="_blank">Edoardo Debenedetti</a>
<a href="https://edoardo.science"><i class="fas fa-globe"></i></a>
<a href="https://github.com/dedeswim"><i class="fab fa-github"></i></a>
<a href="https://twitter.com/edoardo_debe"><i class="fab fa-twitter"></i></a>
</li>
</ul>
</div>
</div>
</section>
</div>
<hr class="bottomrule" />
<footer>
<small>© 2021, RobustBench;
<a href="https://icons8.com/icon/100413/access">Icons from Icons8</a></small>
</footer>
<script>
// When the user scrolls the page, execute myFunction
window.onscroll = function () {
myFunction();
};
// Get the navbar
var navbar = document.getElementById("navbar");
// Get the offset position of the navbar
var sticky = navbar.offsetTop;
// Add the sticky class to the navbar when you reach its scroll position. Remove "sticky" when you leave the scroll position
function myFunction() {
if (window.pageYOffset >= sticky) {
navbar.classList.add("sticky");
} else {
navbar.classList.remove("sticky");
}
}
</script>
<script>
$("#div_cifar10_Linf").load("./cifar10/Linf.html");
$("#div_cifar10_L2").load("./cifar10/L2.html");
$("#div_cifar10_corruptions").load("./cifar10/corruptions.html");
$("#div_cifar100_Linf").load("./cifar100/Linf.html");
$("#div_cifar100_corruptions").load("./cifar100/corruptions.html");
$("#div_imagenet_Linf").load("./imagenet/Linf.html");
$("#div_imagenet_corruptions").load("./imagenet/corruptions.html");
</script>
</body>