-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathDehazeFlow.html
136 lines (134 loc) · 5.92 KB
/
DehazeFlow.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
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<title>DehazeFlow</title>
<link rel="stylesheet" type="text/css" href="assets/scripts/bulma.min.css">
<link rel="stylesheet" type="text/css" href="assets/scripts/theme.css">
<link rel="stylesheet" type="text/css" href="https://cdn.bootcdn.net/ajax/libs/font-awesome/4.7.0/css/font-awesome.min.css">
</head>
<body>
<section class="hero is-light" style="">
<div class="hero-body" style="padding-top: 50px;">
<div class="container" style="text-align: center;margin-bottom:5px;">
<h1 class="title">
DehazeFlow: Multi-scale Conditional Flow Network <br> for Single Image Dehazing
</h1>
<div class="author">Hongyu Li<sup>1</sup></div>
<div class="author">Jia Li<sup>1,2,3*</sup></div>
<div class="author">Dong Zhao<sup>1*</sup>*</div>
<div class="author">Long Xu<sup>2,3</sup></div>
<div class="group">
<a href="http://cvteam.net/">CVTEAM</a>
</div>
<div class="aff">
<p><sup>1</sup>State Key Laboratory of Virtual Reality Technology and Systems, SCSE, Beihang University, Beijing, China</p>
<p><sup>2</sup>Key Laboratory of Solar Activity, National Astronomical Observatories, Chinese Academy of Sciences, Beijing, China</p>
<p><sup>3</sup>Peng Cheng Laboratory, Shenzhen, China</p>
</div>
<div class="con">
<p style="font-size: 24px; margin-top:5px; margin-bottom: 15px;">
ACM MM 2021
</p>
</div>
<div class="columns">
<div class="column"></div>
<div class="column"></div>
<div class="column">
<!-- <a href="https://arxiv.org/abs/2008.04031" target="_blank">
<p class="link">Paper</p>
</a> -->
<p class="link">Paper</p>
</div>
<div class="column">
<p class="link">Code</p>
</div>
<div class="column"></div>
<div class="column"></div>
</div>
</div>
</div>
</section>
<div style="text-align: center;">
<div class="container" style="max-width:850px">
<div style="text-align: center;">
<img src="assets/DehazeFlow/network.png" class="centerImage">
</div>
</div>
<div class="head_cap">
<p style="color:gray;">
The framework of DehazeFlow
</p>
</div>
</div>
<section class="hero">
<div class="hero-body">
<div class="container" style="max-width: 800px" >
<h1 style="">Abstract</h1>
<p style="text-align: justify; font-size: 17px;">
Single image dehazing is a crucial and preliminary task for many computer vision applications, making progress with deep learning.
The dehazing task is an ill-posed problem since the haze in the image leads to the loss of information. Thus, there are multiple feasible
solutions for image restoration of a hazy image. Most existing methods learn a deterministic one-to-one mapping between a hazy image and
its ground-truth, which ignores the ill-posedness of the dehazing task. To solve this problem, we propose DehazeFlow, a novel single image
dehazing framework based on conditional normalizing flow. Our method learns the conditional distribution of haze-free images given a hazy
image, enabling the model to sample multiple dehazed results. Furthermore, we propose an attention-based coupling layer to enhance the
expression ability of a single flow step, which converts natural images into latent space and fuses features of paired data. These designs
enable our model to achieve state-of-the-art performance while considering the ill-posedness of the task. We carry out sufficient experiments
on both synthetic datasets and real-world hazy images to illustrate the effectiveness of our method. The extensive experiments indicate
that DehazeFlow surpasses the state-of-the-art methods in terms of PSNR, SSIM, LPIPS, and subjective visual effects.
</p>
</div>
</div>
</section>
<section class="hero is-light" style="background-color:#FFFFFF;">
<div class="hero-body">
<div class="container" style="max-width:800px;margin-bottom:20px;">
<h1>
Qualitative Comparison
</h1>
</div>
<div class="container" style="max-width:800px">
<div style="text-align: center;">
<img src="assets/DehazeFlow/comparison.png" class="centerImage">
</div>
</div>
</div>
</section>
<section class="hero is-light" style="background-color:#FFFFFF;">
<div class="hero-body">
<div class="container" style="max-width:800px;margin-bottom:20px;">
<h1>
Quantitative Comparison
</h1>
</div>
<div class="container" style="max-width:800px">
<div style="text-align: center;">
<img src="assets/DehazeFlow/metrics.png" class="centerImage">
</div>
</div>
</div>
</section>
<section class="hero" style="padding-top:0px;">
<div class="hero-body">
<div class="container" style="max-width:800px;">
<div class="card">
<header class="card-header">
<p class="card-header-title">
BibTex Citation
</p>
<a class="card-header-icon button-clipboard" style="border:0px; background: inherit;" data-clipboard-target="#bibtex-info" >
<i class="fa fa-copy" height="20px"></i>
</a>
</header>
<div class="card-content">
<pre style="background-color:inherit;padding: 0px;" id="bibtex-info">
To be updated...
</pre>
</div>
</section>
<script type="text/javascript" src="assets/scripts/clipboard.min.js"></script>
<script>
new ClipboardJS('.button-clipboard');
</script>
</body>
</html>