-
Notifications
You must be signed in to change notification settings - Fork 0
/
explainable-virus-quantification.html
652 lines (572 loc) · 33.4 KB
/
explainable-virus-quantification.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
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<!-- Meta tags for social media banners, these should be filled in appropriatly as they are your "business card" -->
<!-- Replace the content tag with appropriate information -->
<meta name="description" content="DESCRIPTION META TAG">
<meta property="og:title" content="DEEP-EM: Explainable Virus Quantification" />
<meta property="og:description"
content="Unlock the power of Deep Learning in Electron Microscopy with the DEEP-EM TOOLBOX standardized workflows for EM image analysis." />
<meta property="og:url" content="URL OF THE WEBSITE" />
<!-- Path to banner image, should be in the path listed below. Optimal dimenssions are 1200X630-->
<meta property="og:image" content="static/image/your_banner_image.png" />
<meta property="og:image:width" content="1200" />
<meta property="og:image:height" content="630" />
<meta name="twitter:title" content="DEEP-EM: Explainable Virus Quantification">
<meta name="twitter:description"
content="Unlock the power of Deep Learning in Electron Microscopy with the DEEP-EM TOOLBOX standardized workflows for EM image analysis.">
<!-- Path to banner image, should be in the path listed below. Optimal dimenssions are 1200X600-->
<meta name="twitter:image" content="static/images/your_twitter_banner_image.png">
<meta name="twitter:card" content="summary_large_image">
<!-- Keywords for your paper to be indexed by-->
<meta name="keywords" content="Deep Learning, Electron Microscopy, Data Analysis, Data Interpretation, Toolbox">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>DEEP-EM: Explainable Virus Quantification</title>
<link rel="icon" type="image/x-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="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="static/css/index.css">
<style>
.hidden {
display: none;
}
.button-55 {
align-self: center;
background-color: #fff;
background-image: none;
background-position: 0 90%;
background-repeat: repeat no-repeat;
background-size: 4px 3px;
border-radius: 15px 225px 255px 15px 15px 255px 225px 15px;
border-style: solid;
border-width: 2px;
box-shadow: rgba(0, 0, 0, .2) 15px 28px 25px -18px;
box-sizing: border-box;
color: #41403e;
cursor: pointer;
display: inline-block;
font-family: Neucha, sans-serif;
font-size: 1rem;
line-height: 23px;
outline: none;
padding: .75rem;
text-decoration: none;
transition: all 235ms ease-in-out;
border-bottom-left-radius: 15px 255px;
border-bottom-right-radius: 225px 15px;
border-top-left-radius: 255px 15px;
border-top-right-radius: 15px 225px;
user-select: none;
-webkit-user-select: none;
touch-action: manipulation;
}
.button-55:hover {
box-shadow: rgba(0, 0, 0, .3) 2px 8px 8px -5px;
transform: translate3d(0, 2px, 0);
}
.button-55:focus {
box-shadow: rgba(0, 0, 0, .3) 2px 8px 4px -6px;
}
.green-background {
background-color: #dde9afff;
/* Green background */
padding: 20px;
/* Padding inside the element */
border-radius: 10px;
/* Rounded corners */
box-shadow: 0 4px 8px rgba(0, 0, 0, 0.2);
/* Subtle shadow for depth */
font-size: 16px;
/* Font size */
margin: 20px 0;
/* Margin outside the element */
}
.red-background {
background-color: #ffaaaaff;
/* Green background */
padding: 20px;
/* Padding inside the element */
border-radius: 10px;
/* Rounded corners */
box-shadow: 0 4px 8px rgba(0, 0, 0, 0.2);
/* Subtle shadow for depth */
font-size: 16px;
/* Font size */
margin: 20px 0;
/* Margin outside the element */
}
.orange-background {
background-color: #ffb380ff;
/* Green background */
padding: 20px;
/* Padding inside the element */
border-radius: 10px;
/* Rounded corners */
box-shadow: 0 4px 8px rgba(0, 0, 0, 0.2);
/* Subtle shadow for depth */
font-size: 16px;
/* Font size */
margin: 20px 0;
/* Margin outside the element */
}
</style>
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script src="https://documentcloud.adobe.com/view-sdk/main.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="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js"></script>
<script>
function toggleVisibility(id_content, id_button) {
var content = document.getElementById(id_content);
var btn = document.getElementById(id_button);
if (content.classList.contains('hidden')) {
content.classList.remove('hidden');
btn.innerHTML = "Show Less"
} else {
content.classList.add('hidden');
btn.innerHTML = "Show More"
}
}
</script>
</head>
<body>
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<!--<img src="static/images/icon.png"
alt="Schematic showing 3 differnt types of task applicable for deep learning. (image to values, image to image & 2D to 3D)" />
-->
<h1 class="title is-1 publication-title">Image to Value(s)</h1>
<!--Add use case category (one of: Image to Value(s), Image to Image, 2D to 3D)-->
<h2 class="title is-1 publication-title">Explainable Virus Quantification</h2>
<!-- Add title of the use case-->
<div class="is-size-5 publication-authors">
<!-- authors -->
<span class="author-block">
<a href="https://viscom.uni-ulm.de/members/hannah-kniesel/" target="_blank">Hannah
Kniesel</a><sup>1</sup>,</span>
<span class="author-block">
<a href="https://viscom.uni-ulm.de/members/tristan-payer/" target="_blank">Tristan
Payer</a><sup>1</sup>,</span>
<span class="author-block">
<a href="https://viscom.uni-ulm.de/members/poonam/" target="_blank">Poonam
Poonam</a><sup>1</sup>,
</span>
<span class="author-block">
<a href="" target="_blank">Tim Bergner</a><sup>2</sup>,
</span>
<span class="author-block">
<a href="https://phermosilla.github.io/" target="_blank">Pedro
Hermosilla</a><sup>3</sup>
</span>
<span class="author-block">
<a href="https://viscom.uni-ulm.de/members/timo-ropinski/" target="_blank">Timo
Ropinski</a><sup>1</sup>,
</span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block"><sup>1</sup>Visual Computing Group, Ulm
University<br><sup>2</sup>Central
Facility for Electron Microscopy, Ulm University<br><sup>3</sup>Computer Vision Lab, TU
Vienna</span>
<!-- <span class="eql-cntrb"><small><br><sup>*</sup>Indicates Equal Contribution</small></span> -->
</div>
<div class="column has-text-centered">
<div class="publication-links">
<!-- Arxiv PDF link
<span class="link-block">
<a href="https://arxiv.org/pdf/<ARXIV PAPER ID>.pdf" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Paper</span>
</a>
</span>
-->
<!-- Supplementary PDF link
<span class="link-block">
<a href="static/pdfs/supplementary_material.pdf" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Supplementary</span>
</a>
</span>-->
<!-- Github link
<span class="link-block">
<a href="https://github.com/YOUR REPO HERE" target="_blank"
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> -->
<!-- ArXiv abstract Link
<span class="link-block">
<a href="https://arxiv.org/abs/<ARXIV PAPER ID>" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>arXiv</span>
</a>
</span>-->
<!--<span class="link-block">
<a href="https:/Link to Notebook" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span>notebook</span>
</a>
</span>-->
<a target="_blank" href="https://lightning.ai/hannah-kniesel/studios/deep-em-toolbox-explainable-virus-quantification">
<img src="https://pl-bolts-doc-images.s3.us-east-2.amazonaws.com/app-2/studio-badge.svg" alt="Open In Studio"/>
</a>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- Teaser image
<section class="hero teaser">
<div class="container is-max-desktop">
<div class="hero-body">
<img src="static/images/tasks.png" alt="Schematic showing 3 differnt types of task applicable for deep learning. (image to values, image to image & 2D to 3D)" />
<h2 class="subtitle has-text-centered">We propose to categorize tasks within the area of EM data analysis into Image to Value(s), Image to Image and 2D to 3D. We do so, based on their specific requirements for implementing a deep learning workflow. For more details, please see our paper.</h2>
</div>
</div>
</section>
End teaser image -->
<!-- Motivation -->
<section class="section hero is-light">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="content has-text-justified">
<p>
Within this image-to-value(s) tasks, we are developing a regression model to quantify HCMV
capsids
and their
maturation stages during secondary envelopment in TEM images.
Secondary envelopment of HCMV is the process by which viral capsids acquire a final envelope
by budding into
cytoplasmic vesicles, a step essential for the production
of infectious progeny [1]. During this process, different maturation stages - naked, budding
or enveloped
capsids - can be observed in TEM images [2].
Quantifying these stages and comparing wild-type viruses with mutants that have defects in
secondary
envelopment can provide valuable insights into the proteins involved
and their role in this critical process [2,3,4].
In our notebook we aim to detect the number of naked, budding and enveloped virus particles
within an input
image.
We use pre-trained model weights to avoid over-fitting on the training data provided. We
also focus on
additional explanatory techniques (GradCAM [5]) to make the model more trustworthy, reliable
and easier to
detect incorrect predictions.
</p>
<p><i>[1] Mettenleiter, Thomas C. "Budding events in herpesvirus morphogenesis." Virus research
106.2 (2004):
167-180.</i></p>
<p><i>[2] Read, Clarissa, et al. "Regulation of human cytomegalovirus secondary envelopment by a
C-terminal
tetralysine motif in pUL71." Journal of Virology 93.13 (2019): 10-1128.</i></p>
<p><i>[3] Cappadona, Ilaria, et al. "Human cytomegalovirus pUL47 modulates tegumentation and
capsid accumulation
at the viral assembly complex." Journal of virology 89.14 (2015): 7314-7328.</i></p>
<p><i>[4] Read, Clarissa, Paul Walther, and Jens von Einem. "Quantitative electron microscopy to
study HCMV
morphogenesis." Human Cytomegaloviruses: Methods and Protocols (2021): 265-289.</i></p>
<p><i>[5] Selvaraju, Ramprasaath R., et al. "Grad-cam: Visual explanations from deep networks
via gradient-based
localization." Proceedings of the IEEE international conference on computer vision.
2017.</i></p>
</div>
</div>
</div>
</div>
</section>
<!-- End Motivation -->
<!--DEEP-EM TOOLBOX Workflow -->
<section class="section hero">
<div class="container is-max-desktop content">
<h2 class="title is-3">DEEP-EM TOOLBOX: Workflow</h2>
<div class="content has-text-justified">
<figure>
<img src="static/images/explainable-virus-quantification/Teaser.png"
alt="Teaser for explainable virus quantification">
<figcaption id="fig:teaser">
For the explainable virus quantification we train a regression model to predict the number of
"naked", "budding" and "enveloped" virus capsids in the input image.
We use GradCAM [5] as an explainable AI technique, to make the model more trustworthy and the
predictions easier to coprehend.
</figcaption>
</figure>
<div class="orange-background">
<h3>Task</h3>
<p>
We categorize the <u>task</u> of this use case into <u>Image to Value(s)</u> as the input will
be an
<abbr title="Electron Microscopy">EM</abbr> image, and the output will be three values defining
the number
of naked, budding, and enveloped HCMV capsids in the input image. Next, we define the model
<u>architecture</u>:
for selecting the <u>backbone</u>, we take into account that we are working with image data,
which opens the
possibility of using <abbr title="Convolutional Neural Network">CNN</abbr> or
<abbr title="Vision Transformer">ViT</abbr> architectures. Given the limited amount of data, we
choose to
use <abbr title="Convolutional Neural Network">CNN</abbr> models. Specifically, we select
ResNet50
[1] because it
allows us to
utilize pretrained <abbr title="Electron Microscopy">EM</abbr> weights from CEM500k
[2].
Regarding the <u>task-specific</u> architecture, we follow the standard of using the backbone as
a feature
encoder and defining a task-specific head with an output dimensionality of three, matching the
number of
values to predict the secondary envelopment stages within the
<abbr title="Transmission Electron Microscopy">TEM</abbr> images.
</p>
</div>
<div class="green-background">
<h3>Data</h3>
<p>
For data <u>acquisition</u>, we make use of pre-existing <u>real data</u> from
[3]
that has already been <u>annotated</u>. However, we encourage the research community to gather
their own datasets
to use our example use case. The <u>annotation type</u> of the data are instance annotations
marking capsid locations
with classes \(C=\{\text{"naked"}, \text{"budding"}, \text{"enveloped"}\}\). For your own data,
we recommend using the
CVAT <u>annotation tool</u>
[4],
as this will allow you to directly plug in your data to run the provided script.
When using your own data, it is also possible to change the number and value of the classes. For more
details, we refer the reader
to our project page and the corresponding notebook. Based on these annotations, we can derive
the number of naked,
budding, and enveloped virus capsids.
</p>
<p>
For data <u>preprocessing</u>, we <u>split</u> the dataset into training, validation, and test
sets, using 20% for testing,
20% for validation, and 60% for training. We choose a relatively large test split, as dataset
sizes in
<abbr title="Electron Microscopy">EM</abbr> are usually small. By doing so, we can ensure that
the model is sufficiently
evaluated to account for potential overfitting to the training data. For <u>resizing</u> we crop
random patches of
224×224 pixels from the full-resolution images, to match the input size of our pretrained
backbone during training.
On the validation and test set we work in a sliding window
fashion. We apply image
<u>normalization</u> by z-score normalization, where we use the mean and standard deviation from
the pretrained dataset.
This is done to match the distribution of the pretrained dataset and better reuse the learned
features from pretraining.
</p>
<p>
To further increase variance in the dataset, we apply standard <u>augmentation</u> techniques
like resizing and rotating
the input images.
</p>
</div>
<div class="red-background">
<h3> Model</h3>
<p>
As mentioned before, we <u>initialize</u> the backbone model with
<abbr title="Electron Microscopy">EM</abbr>-specific pretrained weights from CEM500k
[2].
We also <u>log</u> important metrics using
<a href="https://wandb.ai/site" target="_blank">Weights & Biases</a>, such as training loss,
validation loss, and
validation mean average error. We visualize a subset of the validation set during training with
model input, output,
prediction, and ground truth. This allows us to quantitatively and qualitatively assess the
training development of the model.
We specifically deploy explainability using GradCAM
[5],
which helps us to sidestep the black-box nature of the model.
</p>
<p>
Lastly, we <u>evaluate</u> the overall performance of the model on the test set. We do this in a
<u>quantitative</u> and <u>qualitative</u> fashion and again use GradCAM as a method of
<u>explaining</u> the model's predictions. This allows us to visualize possible biases in the
trained model.
</p>
</div>
<p>[1] He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In
Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 770-778).</p>
<p>[2] Conrad, Ryan, and Kedar Narayan. "CEM500K, a large-scale heterogeneous unlabeled cellular
electron
microscopy image dataset for deep learning." Elife 10 (2021): e65894.</p>
<p>[3] Shaga Devan, Kavitha, et al. "Improved automatic detection of herpesvirus secondary envelopment
stages in electron microscopy by augmenting training data with synthetic labelled images generated
by a generative adversarial network." Cellular Microbiology 23.2 (2021): e13280.</p>
<p>[4] B. Sekachev et al., Opencv/cvat: V1.1.0, version v1.1.0, Aug. 2020. doi: 10 . 5281 / zenodo .
4009388. [Online]. Available: https://doi.org/10.5281/zenodo.4009388.</p>
<p>[5] Selvaraju, Ramprasaath R., et al. "Grad-cam: Visual explanations from deep networks via
gradient-based localization." Proceedings of the IEEE international conference on computer vision.
2017.</p>
</div>
</div>
</section>
<!--End DEEP-EM TOOLBOX Workflow-->
<!--Use your own data -->
<section class="section hero is-light">
<div class="container is-max-desktop content">
<h2 class="title is-3">Use Your Own Data</h2>
<div class="content has-text-justified">
<p>
Here we explain how you need to preprocess and annotate your data to apply the model for your use
case. We require that all data for training, testing and validation is annotated
and collected and then uploaded to the pytorch lightning studio.</p>
<h3>Data Structuring</h3>
<p>We require you to put all your data into a single folder. </p>
<p>The data format of the micrographs should be <code>.tif</code></p>
<p>Follow the instructions below for annotating the data. This will lead to a single <code>.xml</code> file
containing all data labels. Please put the resulting <code>.xml</code> file into the same folder, where you store
all micrographs. Finally, the annotation file needs to be named <code>"annotation_file.xml"</code></p>
<p></p>
<h3>Data Annotation</h3>
<p>
For data annotation we recomment using the <a href="https://www.cvat.ai/">CVAT</a> tool. It is
free to use
but requires a user account.
A quick user guide can be found <a href="https://docs.cvat.ai/docs/getting_started/">here</a>.
Multiple instruction videos can also be found on the <a
href="https://www.youtube.com/@LearnOpenCV">LearnOpenCV youtube channel</a> (for example <a
href="https://www.youtube.com/watch?v=yxX_0-zr-2U">this</a>).
For annotating the location labels within this use case we implement the use of point
annotations.
We export the data using the <i>CVAT for images 1.1</i> format.
</p>
<button id="togglebtn-quan-ann" class="button-55"
onclick="toggleVisibility('content-quan-ann', 'togglebtn-quan-ann')">Show
More</button>
<div id="content-quan-ann" class="hidden">
<p></p>
<figure>
<img src="static/images/explainable-virus-quantification/CVAT-1.png" alt="CVAT task 1">
</figure>
<p>When starting CVAT, you first need to create a new task. You can give it a name, add annotation
types and upload your data.</p>
<figure>
<img src="static/images/explainable-virus-quantification/CVAT-2.png" alt="CVAT task 2">
</figure>
<p>Next, click on the <b>Add label</b> button. Name it based on the classes you want to annotate. In
our case these are "naked", "budding", "enveloped".
As annotation type choose <b>Points</b>. You should also pick a color, as this will simplify the
annotation process. For adding new labels click <b>Continue</b>.
Once you added all nessecary labels click <b>Cancel</b>.
</p>
<figure>
<img src="static/images/explainable-virus-quantification/CVAT-3.png" alt="CVAT task 3">
</figure>
<p>Now you can upload the data you wish to annotate. Finally, click <b>Submit & Open</b> to continue
with the annotation of the uploaded data.
</p>
<figure>
<img src="static/images/explainable-virus-quantification/CVAT-4.png" alt="CVAT job">
</figure>
<p>This will open following view. Click on the <b>job</b> (in this view the <b>Job #1150022</b>) to
start the annotation job.
</p>
<figure>
<img src="static/images/explainable-virus-quantification/CVAT-5.png" alt="CVAT annotate">
</figure>
<p>To then annotate your data, select the <b>Draw new points</b> tool. Select the <b>Label</b> you
wish to annotate from the dropdown menue. Then click
<b>Shape</b> to annotate
individual virus capsids with the label class. (<b>Track</b> will allow you to place annotations
over multiple frames, which is helpful when annotating videos, tomograms or similar).
You can use the arrows on the top middle to navigate through all of your data and to see your
annotation progress.
</p>
<figure>
<img src="static/images/explainable-virus-quantification/CVAT-6.png" alt="CVAT export 1">
</figure>
<p>Once you are done annotating data, click on the <b>Menu</b> and select <b>Export job dataset</b>.
</p>
<figure>
<img src="static/images/explainable-virus-quantification/CVAT-7.png" alt="CVAT export 2">
</figure>
<p>During export select the <b>CVAT for Images 1.1</b> format and give the folder a name. It will
prepare the dataset for download.
</p>
<figure>
<img src="static/images/explainable-virus-quantification/CVAT-8.png" alt="CVAT download">
</figure>
<p>In the horizontal menu bar at the top go to <b>Requests</b>. It will show a request <b>Export
Annotations</b>. On the right of this request click on the three dots to download your
annotated data. And you are done.
</p>
</div>
</div>
</div>
</section>
<!--End use your own data -->
<!--Links -->
<section class="section hero">
<div class="container is-max-desktop content">
<h2 class="title is-3">Contact</h2>
<div class="content has-text-justified">
<p>If you have any questions regarding this use case, please do not hesitate to contact <a
href="mailto:deep_em@tristan.anonaddy.com">Tristan Payer</a></p> <!--Add contact info here-->
</div>
</div>
</section>
<!--End Links -->
<!--BibTex citation -->
<section class="section is-light" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title">BibTeX</h2>
<pre><code>BibTex Code Here</code></pre>
</div>
</section>
<!--End BibTex citation -->
<!-- Footer -->
<footer class="footer">
<div class="container">
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p>
This page was built using the <a
href="https://github.com/eliahuhorwitz/Academic-project-page-template"
target="_blank">Academic Project Page Template</a> which was adopted from the <a
href="https://nerfies.github.io" target="_blank">Nerfies</a> project page.
You are free to borrow the of this website, we just ask that you link back to this page in
the footer.
<br> This website is licensed under a <a rel="license"
href="http://creativecommons.org/licenses/by-sa/4.0/" target="_blank">Creative
Commons Attribution-ShareAlike 4.0 International License</a>.
</p>
</div>
</div>
</div>
</div>
</footer>
<!-- End footer -->
<!-- Statcounter tracking code -->
<!-- You can add a tracker to track page visits by creating an account at statcounter.com -->
<!-- End of Statcounter Code -->
</body>
</html>