-
Notifications
You must be signed in to change notification settings - Fork 3
/
index.html
323 lines (286 loc) · 12.1 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
<!-- Thanks to url=http://www.cs.cmu.edu/~dfouhey/3DP/index.html -->
<!DOCTYPE HTML>
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<link rel="StyleSheet" href="assets/style.css" type="text/css" media="all">
<title>IDEAL</title>
<script src="https://kit.fontawesome.com/c444c87c0c.js" crossorigin="anonymous"></script>
<script src="https://code.iconify.design/iconify-icon/1.0.2/iconify-icon.min.js"></script>
<script type="text/javascript" async="" src="assets/ga.js"></script>
<script type="text/javascript">
</script>
<style>
table.bordered,
th.bordered,
td.bordered {
border: 1px solid black;
border-collapse: collapse;
padding: 15px;
}
.center {
margin-left: auto;
margin-right: auto;
}
</style>
<!-- bibliographic tags -->
<meta name="citation_title" content="IDEAL: Improved DEnse locAL Contrastive Learning for Semi-Supervised Medical Image Segmentation" />
<meta name="citation_author" content="Basak, Hritam" />
<meta name="citation_author" content="Chattopadhyay, Soumitri" />
<meta name="citation_author" content="Kundu, Rohit" />
<meta name="citation_author" content="Nag, Sayan" />
<meta name="citation_author" content="Mallipeddi, Rammohan" />
<meta name="citation_publication_date" content="20XX" />
<meta name="citation_conference_title" content="XXXX" />
<meta name="citation_pdf_url" content="https://arxiv.org/abs/2210.15075" />
<style type="text/css">
#primarycontent h1 {
font-variant: small-caps;
}
#primarycontent h3 {}
#primarycontent teasertext {
text-align: center;
}
#primarycontent p {
text-align: center;
}
#primarycontent {
text-align: justify;
}
#primarycontent p {
text-align: justify;
}
#primarycontent p iframe {
text-align: center;
}
#avatar {
border-radius: 50%;
}
</style>
<script type="text/javascript">
function togglevis(elid) {
el = document.getElementById(elid);
aelid = elid + "a";
ael = document.getElementById(aelid);
if (el.style.display == 'none') {
el.style.display = 'inline-table';
ael.innerHTML = "[Hide BibTex]";
} else {
el.style.display = 'none';
ael.innerHTML = "[Show BibTex]";
}
}
</script>
</head>
<body>
<div id="primarycontent">
<h1 align="center" itemprop="name"><strong>
IDEAL<i style="color:#03C04A">✓</i>: Improved DEnse locAL Contrastive Learning for Semi-Supervised Medical Image Segmentation
</strong></h1>
<table id="authors" style="margin:auto;">
<tr>
<td></td> <!-- For some reason it scales up the first td.. so adding a dummy td -->
<td>
<a href="https://hritam-98.github.io/" target="_blank">Hritam Basak*<sup>1</sup></a>
</td>
<td>
<a href="https://soumitri2001.github.io/" target="_blank">Soumitri Chattopadhyay*<sup>2</sup></a>
</td>
<td>
<a href="https://rohit-kundu.github.io/" target="_blank">Rohit Kundu*<sup>3</sup></a>
</td>
<td>
<a href="https://sayannag.github.io/" target="_blank">Sayan Nag*<sup>4</sup></a>
</td>
<td>
<a href="https://sites.google.com/site/rammohanmallipeddi/" target="_blank">Rammohan Mallipeddi<sup>5</sup></a>
</td>
</tr>
</table>
<table id="affliates" style="margin:auto;">
<tr>
<td></td> <!-- For some reason it scales up the first td.. so adding a dummy td -->
<td>
<p><sup>1</sup>Stony Brook University</p>
</td>
<td>
<p><sup>2</sup>Jadavpur University</p>
</td>
<td>
<p> <sup>3</sup>University of California, Riverside</p>
</td>
<td>
<p> <sup>4</sup>University of Toronto</p>
</td>
<td>
<p> <sup>5</sup>Kyungpook National University</p>
</td>
</tr>
</table>
<table id="navigate" style="margin:auto;">
<tr>
<td>
<iconify-icon icon="bi:file-earmark-pdf-fill"></iconify-icon>
<a href="https://arxiv.org/abs/2210.15075" target="_blank"> arxiv</a>
</td>
<td>
<iconify-icon icon="octicon:mark-github-16"></iconify-icon>
<a href="https://github.com/Rohit-Kundu/IDEAL-ICASSP23" target="_blank"> GitHub</a>
</td>
<td>
<iconify-icon icon="bxs:quote-left"></iconify-icon>
<a href="assets/bib.txt">bibtex</a>
</td>
</tr>
</table>
<table id="news" style="margin:auto;">
<tr>
<td>
<p> <b style="color:crimson">Published</b> at the <a href="https://2023.ieeeicassp.org/" target="_blank">48<sup>th</sup> IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2023) </a></p>
</td>
</tr>
</table>
<h3>Overview</h3>
<table class="results" align="center">
<tr>
<td align="center">
<img src="assets/mainfig.png" width="80%" /></a>
</td>
</tr>
<tr></tr>
<tr></tr>
<tr></tr>
<tr>
<td class="credits" align="justify">
An overview of the proposed <strong>IDEAL</strong> (<b>I</b>mproved <b>DE</b>nse loc<b>AL</b> Contrastive Learning for Semi-Supervised Medical Image Segmentation) framework: (a) <i>Pre-training</i>- <i>x<sub>q</sub></i> and <i>x<sub>k</sub></i> are the query and key images, <i>E</i> and
<i>G</i> represent the encoder and projection head, respectively. The projection head employs a 1x1 convolution layer instead of a
traditional MLP for dense feature extraction, resulting in better local clustering of features. (b) <i>Fine-tuning</i>- Two perturbed
branches with the same input are employed. <i>E</i>(<i>θ</i>) is the shared encoder initialized similarly for both streams, <i>D</i>(<i>θ</i><sub>1</sub>) and <i>D</i>(<i>θ</i><sub>2</sub>)
represent two different decoder architectures; <i>p</i><sub>1</sub> and <i>p</i><sub>2</sub> are the predicted output segmentation maps which are thresholded to
obtain <i>y</i><sub>1</sub> and <i>y</i><sub>2</sub> respectively. <i>y</i><sub>1</sub> backpropagates through the second stream and <i>y</i><sub>2</sub> backpropagates through the first stream
enforcing cross-consistency in segmentation.
</td>
</tr>
<tr>
</tr>
</table>
<h3>Main Results</h3>
<table class="results" align="center">
<tr>
<td align="center">
<img src="assets/table1.png" width="50%" /></a>
</td>
</tr>
<tr></tr>
<tr></tr>
<tr></tr>
<tr>
<td class="credits" align="justify">
Results obtained by the IDEAL framework with
varying amounts of labeled data on the ACDC and MMWHS
datasets. ‘<i>L</i>’ represents the amount of labeled data used.
</td>
</tr>
<tr>
<td align="center">
<img src="assets/table2.png" width="50%" /></a>
</td>
</tr>
<tr></tr>
<tr></tr>
<tr></tr>
<tr>
<td class="credits" align="justify">
Performance Comparison (DSC scores) of the proposed
IDEAL framework with SoTA methods in the literature
on the ACDC and MMWHS datasets.
</td>
</tr>
<tr>
<td align="center">
<img src="assets/segoutputs.png" width="50%" /></a>
</td>
</tr>
<tr></tr>
<tr></tr>
<tr></tr>
<tr>
<td class="credits" align="justify">
Visual comparison of our results with SoTA methods
and ground truth, thus qualitatively validating the superiority
of IDEAL in terms of segmentation performance.
</td>
</tr>
</table>
<h3>People</h3>
<table id="people" style="margin:auto;">
<tr>
<td></td> <!-- For some reason it scales up the first td.. so adding a dummy td -->
<td>
<img src="assets/authors/hritam.jpg" /><br />
<a href="https://hritam-98.github.io/" target="_blank">Hritam Basak</a>
</td>
<td>
<img src="assets/authors/soumitri.jpg" /><br />
<a href="https://soumitri2001.github.io/" target="_blank">Soumitri Chattopadhyay</a>
</td>
<td>
<img src="assets/authors/rohit.jpg" /><br />
<a href="https://rohit-kundu.github.io/" target="_blank">Rohit Kundu</a>
</td>
<td>
<img src="assets/authors/sayan.jpg" /><br />
<a href="https://sayannag.github.io/" target="_blank">Sayan Nag</a>
</td>
<td>
<img src="assets/authors/rm.jpg" /><br />
<a href="https://sites.google.com/site/rammohanmallipeddi/" target="_blank">Rammohan Mallipeddi</a>
</td>
</tr>
</table>
<h3>Acknowledgement</h3>
<table class="results" align="center">
<tr></tr>
<tr></tr>
<tr>
<td class="credits" align="justify">
This work was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2021R1I1A3049810).
</td>
</tr>
<tr>
</tr>
</table>
<h3>Paper</h3>
<table id="paper" class="center">
<tr></tr>
<tr>
<td>
<a href="https://doi.org/10.1109/ICASSP49357.2023.10094869"><img style="box-shadow: 5px 5px 2px #888888; margin: 10px"
src="assets/paper-screenshot.png" width="150px" /></a>
</td>
<td></td>
<td>
Hritam Basak*, Soumitri Chattopadhyay*, Rohit Kundu*, Sayan Nag*, Rammohan Mallipeddi<br />
<a href="https://doi.org/10.1109/ICASSP49357.2023.10094869">IDEAL: Improved DEnse locAL Contrastive Learning for Semi-Supervised Medical Image Segmentation</a><br />
Paper <br />
[<a href="https://arxiv.org/abs/2210.15075">arXiv</a>]
[<a href="https://github.com/Rohit-Kundu/IDEAL-ICASSP23">code</a>]
[<a href="assets/bib.txt" id="bibtex">bibtex</a>] <br /> <br />
</table>
<table class="bibtex" style="display:none" id="basak2022ideal">
<tr>
<td>
<pre>
@article{basak2022ideal,
title={IDEAL: Improved DEnse locAL Contrastive Learning for Semi-Supervised Medical Image Segmentation},
author={Basak, Hritam and Chattopadhyay, Soumitri and Kundu, Rohit and Nag, Sayan and Mallipeddi, Rammohan},
journal={arXiv preprint arXiv:2210.15075},
year={2022}
}
</pre>
</td>
</tr>
</table>
</div>
</body>
</html>