iCVTEAM.github.io/DASNet.html at master · iCVTEAM/iCVTEAM.github.io · GitHub

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
<!DOCTYPE html>
<html>
  <head>
    <meta charset="utf-8">
    <title>DASNet</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">
            Is Depth Really Necessary for Salient Object Detection?
          </h1>

          <div class="author">Jiawei Zhao</div>
          <div class="author">Yifan Zhao</div>
          <div class="author">Jia Li</div>
          <div class="author">Xiaowu Chen</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</p>
          </div>
          <div class="con">
            <p  style="font-size: 24px; margin-top:5px; margin-bottom: 15px;">
            ACMMM 2020
            </p>
          </div>
          <div class="columns">
            <div class="column"></div>
            <div class="column"></div>
            <div class="column">
              <a href="https://arxiv.org/abs/2006.00269" target="_blank">
                <p class="link">Paper</p>
              </a>
            </div>
            <div class="column">
              <a href="https://github.com/iCVTEAM/RGBDSOD/" target="_blank">
                <p class="link">Code</p>
              </a>
            </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/RGBDSOD/head.png" class="centerImage">
        </div>
      </div>
      <div class="head_cap">
        <p style="color:gray;">
          The overall architecture of our model.
        </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;">
            Salient object detection (SOD) is a crucial and preliminary task for
            many computer vision applications, which have made progress with
            deep CNNs. Most of the existing methods mainly rely on the RGB
            information to distinguish the salient objects, which faces difficulties
            in some complex scenarios. To solve this, many recent RGBD-based
            networks are proposed by adopting the depth map as an independent
            input and fuse the features with RGB information. Taking the
            advantages of RGB and RGBD methods, we propose a novel depth-aware
            salient object detection framework, which has following superior designs:
            1) It only takes the depth information as training data while
            only relies on RGB information in the testing phase.
            2) It comprehensively optimizes SOD features with multi-level depth-aware regularizations.
            3) The depth information also serves as error-weighted map to correct the segmentation process.
            With these insightful designs combined, we make the first attempt in realizing an unified
            depth-aware framework with only RGB information as input for
            inference, which not only surpasses the state-of-the-art performance
            on five public RGB SOD benchmarks, but also surpasses the RGBD-based
            methods on five benchmarks by a large margin, while adopting
            less information and implementation light-weighted. The code and
            model will be publicly available.
          </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 comparisons
          </h1>
        </div>
        <div class="container" style="max-width:800px">
          <div style="text-align: center;">
            <img src="assets/RGBDSOD/comp.png" class="centerImage">
          </div>
        </div>
      </div>
    </section>
  </body>
</html>