-
Notifications
You must be signed in to change notification settings - Fork 2
Expand file tree
/
Copy pathICCM.html
More file actions
132 lines (130 loc) · 5.91 KB
/
ICCM.html
File metadata and controls
132 lines (130 loc) · 5.91 KB
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
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<title>ICCM</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">
Informative and Consistent Correspondence Mining for Cross-Domain Weakly Supervised Object Detection
</h1>
<div class="author">Luwei Hou*<sup>1,3</sup></div>
<div class="author">Yu Zhang*<sup>3</sup></div>
<div class="author">Kui Fu<sup>1</sup></div>
<div class="author">Jia Li<sup>1,2</sup></div>
<div class="group">
<a href="http://cvteam.net/">CVTEAM</a>
</div>
<div class="aff">
<p><sup>1</sup>1State Key Laboratory of Virtual Reality Technology and Systems, School of Computer Science and Engineering, Beihang University, Beijing, China</p>
<p><sup>2</sup>Pengcheng Laboratory, Shenzhen, China</p>
<p><sup>3</sup>SenseTime Reasearch</p>
</div>
<div class="con">
<p style="font-size: 24px; margin-top:5px; margin-bottom: 15px;">
CVPR 2021
</p>
</div>
<div class="columns">
<div class="column"></div>
<div class="column"></div>
<div class="column">
<a href="https://openaccess.thecvf.com/content/CVPR2021/papers/Hou_Informative_and_Consistent_Correspondence_Mining_for_Cross-Domain_Weakly_Supervised_Object_CVPR_2021_paper.pdf" target="_blank">
<p class="link">Paper</p>
</a>
</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/ICCM/pipeline.png" class="centerImage">
</div>
</div>
<div class="head_cap">
<p style="color:gray;">
The pipeline of approach
</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;">
Cross-domain weakly supervised object detection aims
to adapt object-level knowledge from a fully labeled source
domain dataset (i.e., with object bounding boxes) to train
object detectors for target domains that are weakly labeled
(i.e., with image-level tags). Instead of domain-level distribution matching, as popularly adopted in the literature, we
propose to learn pixel-wise cross-domain correspondences
for more precise knowledge transfer. It is realized through
a novel cross-domain co-attention scheme trained as region
competition. In this scheme, the cross-domain correspondence module seeks for informative features on the target
domain image, which if warped to the source domain image, could best explain its annotations. Meanwhile, a collaborative mask generator competes to mask out the relevant target image region to make the remaining features
uninformative. Such competitive learning strives to correlate the full foreground in cross-domain image pairs, revealing the accurate object extent in target domain. To alleviate the ambiguity of inter-domain correspondence learning,
a domain-cycle consistency regularizer is further proposed
to leverage the more reliable intra-domain correspondence.
The proposed approach achieves consistent improvements
over existing approaches by a considerable margin, demonstrated by the experiments on various datasets.
</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/ICCM/results.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">@InProceedings{Hou_2021_CVPR,
title = {Informative and Consistent Correspondence Mining for Cross-Domain Weakly Supervised Object Detection},
author = {Hou, Luwei and Zhang, Yu and Fu, Kui and Li, Jia},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
pages = {9929-9938}
month = {June},
year = {2021},
}
</pre>
</div>
</section>
<script type="text/javascript" src="assets/scripts/clipboard.min.js"></script>
<script>
new ClipboardJS('.button-clipboard');
</script>
</body>
</html>