forked from sudoparsa/Decompose-and-Compose
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
314 lines (260 loc) · 14.9 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
<!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="Decompose and Compose">
<meta property="og:title" content="Decompose and Compose"/>
<meta property="og:description" content="Decompose and Compose"/>
<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="Decompose and Compose">
<meta name="twitter:description" content="CVPR 2024">
<!-- 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="KEYWORDS SHOULD BE PLACED HERE">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Decompose and Compose: A Compositional Approach to Mitigating Spurious Correlation</title>
<link rel="icon" type="image/x-icon" href="static/images/favicon.ico">
<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">
<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>
</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">
<h1 class="title is-1 publication-title">Decompose and Compose: A Compositional Approach to Mitigating Spurious Correlation</h1>
<div class="is-size-5 publication-authors">
<!-- Paper authors -->
<span class="author-block">
<a href="https://scholar.google.com/citations?view_op=list_works&hl=en&user=hWWXh3wAAAAJ" target="_blank">Fahimeh Hosseini Noohdani</a>,</span>
<span class="author-block">
<a href="https://scholar.google.com/citations?user=4QnHQqgAAAAJ&hl=en&oi=sra" target="_blank">Parsa Hosseini</a><sup>*</sup>,</span>
<span class="author-block">
<a href="https://scholar.google.com.au/citations?user=7EtxHvgAAAAJ&hl=en" target="_blank">Arian YazdanParast</a><sup>*</sup>,
</span>
<span class="author-block">
<a href="https://scholar.google.com/citations?user=oa9KkfkAAAAJ&hl=en" target="_blank">Hamidreza Yaghoubi Araghi</a>,
</span>
<span class="author-block">
<a href="https://sharif.edu/~soleymani/" target="_blank">Mahdieh Soleymani Baghshah</a>
</span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block">Sharif University of Technology<br>CVPR 2024</span>
<span class="eql-cntrb"><small><br><sup>*</sup>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://openaccess.thecvf.com/content/CVPR2024/html/Noohdani_Decompose-and-Compose_A_Compositional_Approach_to_Mitigating_Spurious_Correlation_CVPR_2024_paper.html" 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/fhn98/DaC" 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/2402.18919" 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>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- Paper abstract -->
<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">
<h2 class="title is-3">Overview</h2>
<div class="content has-text-justified">
<p>
While standard Empirical Risk Minimization (ERM) training is proven effective for image classification on in-distribution data, it fails to perform well on out-of-distribution samples. One of the main sources of distribution shift for image classification is the compositional nature of images. Specifically, in addition to the main object or component(s) determining the label, some other image components usually exist, which may lead to the shift of input distribution between train and test environments. More importantly, these components may have spurious correlations with the label. To address this issue, we propose Decompose-and-Compose (DaC), which improves robustness to correlation shift by a compositional approach based on combining elements of images. Based on our observations, models trained with ERM usually highly attend to either the causal components or the components having a high spurious correlation with the label (especially in datapoints on which models have a high confidence). In fact, according to the amount of spurious correlation and the easiness of classification based on the causal or non-causal components, the model usually attends to one of these more (on samples with high confidence). Following this, we first try to identify the causal components of images using class activation maps of models trained with ERM. Afterward, we intervene on images by combining them and retraining the model on the augmented data, including the counterfactual ones. Along with its high interpretability, this work proposes a group-balancing method by intervening on images without requiring group labels or information regarding the spurious features during training. The method has an overall better worst group accuracy compared to previous methods with the same amount of supervision on the group labels in correlation shift.
</p>
</div>
</div>
</div>
</div>
</section>
<!-- End paper abstract -->
<!-- Image carousel -->
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-five-fifths">
<h2 class="title is-3"> Identifing the causal component </h2>
</div>
</div>
<div class="columns is-centered has-text-centered">
<div class="column is-five-fifths">
<div style="text-align: left;">
<img id="model" src="static/images/Causal.png" alt="Causal Model Image" style="width: 100%; height: auto;">
<h3 class="subtitle" style="font-family: 'Times New Roman', Times, serif;">
<p><b>Figure 1. Behaviour of a model trained with standard ERM in different datasets. Based on the easiness of inferring the label from the
causal or non-causal parts across the whole dataset, the model attends more to one of them, this behaviour is more evident on samples on
which the model has a low loss. (a), (b) Average xGradCAM score of Cifar10 (causal) and and MNIST (non-causal) pixels in four loss
quantiles of the Dominos train set. The model generally attends more to the non-causal parts, and as the loss decreases, the non-causal
attention increases. (c), (d) Average xGradCAM score of foreground (causal) and background (non-causal) pixels in four loss quantiles of
the Waterbirds train set. The model generally pays attention to the causal parts, and as the loss decreases, the causal attention increases.</b></p>
</h3>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-five-fifths">
<h2 class="title is-3"> Intervening on images </h2>
</div>
</div>
<div class="columns is-centered has-text-centered">
<div class="column is-five-fifths">
<div style="text-align: left;">
<img id="model" src="static/images/DAC - Copy.png" alt="Causal Model Image" style="width: 100%; height: auto;">
<h3 class="subtitle" style="font-family: 'Times New Roman', Times, serif; justify-content:end;">
<p><b>Figure 2. An overview of our DaC method.First, we choose one of these assumptions: the mask or its inverse captures the core parts more effectively. Then, we select low-loss images and merge images with different labels. </b></p>
</h3>
</div>
</div>
</div>
</div>
</section>
<section class="section" style="background-color:white;">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Masking process</h2>
</div>
</div>
</div>
<div class="columns is-centered has-text-centered">
<div style="width:90%;">
<div style="float:left; width:47.1%; border: 0px solid black;">
<h2 class="title is-5">Adaptive masking</h2>
<div class="content has-text-justified">
When gradually masking an image, the less predictive parts are masked out first, causing minimal impact on the loss. However, as more of the image, including the predictive parts, is masked, the loss increases rapidly because the model struggles to predict the label. Thus, the optimal masking portion is just before this rapid increase in loss.
</div>
<div class="column is-five-fifths">
<div class="columns is-centered">
<img id="modulated_training" width="75%" src="static/images/Adaptive.png">
</div>
</div>
</div>
<div style=" float:right; width:47.8%; border: 0px solid black;">
<h2 class="title is-5"> Mask and Combine</h2>
<div class="content has-text-justified">
The two input images with different labels are masked using Adaptive masking. Then, the selected part of one image and the masked parts of the other image are combined, with the remaining gaps filled using the mean value of the batch. The new combined image retains the label of the first image and is used for training the last layer of the model.
</div>
<div class="column is-five-fifths">
<div class="columns is-centered">
<img id="modulated_training" width="75%" src="static/images/MaskCombine.png">
</div>
</div>
</div>
</div>
</div>
</section>
<!-- End image carousel -->
<!-- Paper poster -->
<section class="hero is-small is-light">
<div class="hero-body">
<div class="container">
<h2 class="title">Poster</h2>
<iframe src="static/pdfs/output_11.pdf" width="100%" height="550">
</iframe>
</div>
</div>
</section>
<!--End paper poster -->
<!--BibTex citation -->
<section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title">BibTeX</h2>
<pre><code>
@InProceedings{Noohdani_2024_CVPR,
author = {Noohdani, Fahimeh Hosseini and Hosseini, Parsa and Parast, Aryan Yazdan and Araghi, Hamidreza Yaghoubi and Baghshah, Mahdieh Soleymani},
title = {Decompose-and-Compose: A Compositional Approach to Mitigating Spurious Correlation},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2024},
pages = {27662-27671}
}
</code></pre>
</div>
</section>
<!--End BibTex citation -->
<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.
<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>
<!-- 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>