Datasets.github.io/index.html at master · yhldrf/Datasets.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
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
<!DOCTYPE html>
<html lang="en">
<head>
  <meta charset="utf-8">
  <title>6D Object Pose Estimation</title>
  <meta name="description" content="challenge: 6D object pose estimation">
  <script type="text/javascript" src="http://cdn.mathjax.org/mathjax/latest/MathJax.js?config=AM_HTMLorMML-full"></script>
  <meta name="viewport" content="width=device-width, initial-scale=1">
  <link href="//fonts.googleapis.com/css?family=Raleway:400,300,600" rel="stylesheet" type="text/css">
  <link rel="stylesheet" href="css/normalize.css">
  <link rel="stylesheet" href="css/skeleton.css">

  <style>
    body {
      font-family: 'Raleway', sans-serif;
      font-size: 18px;
      line-height: 1.7;
    }

    .content {
      margin-top: 5rem;
      margin-bottom: 5rem;
    }

    .logo {
      height: 100px;
      display: block;
      margin: 0 auto 2rem auto;
    }

    h2.title {
      text-align: center;
      font-size: 4rem;
      font-weight: 700;
      margin-bottom: 2rem;
    }

    h3 {
      font-size: 2rem;
      font-weight: 600;
      margin-top: 2.5rem;
      margin-bottom: 1rem;
    }

    h5 {
      font-size: 1.5rem;
      font-weight: 500;
      margin-top: 2rem;
      margin-bottom: 0.5rem;
    }

    p {
      margin-bottom: 1rem;
    }

    .button-full {
      width: 100%;
      margin: 1rem 0;
    }

    .image-row {
      display: flex;
      flex-wrap: wrap;
      gap: 1rem;
      justify-content: center;
      margin: 1rem 0 2rem 0;
    }

    .image-column {
      flex: 1 1 45%;
      min-width: 250px;
    }

    .image-column img {
      width: 100%;
      height: auto;
      max-height: 350px;
      object-fit: contain;
    }

    .center-text {
      text-align: center;
      font-size: 18px;
      margin-top: 0.5rem;
      margin-bottom: 1rem;
    }

    ul {
      margin-left: 1.5rem;
    }
  </style>
</head>

<body>
  <div class="container">
    <section class="content">
      <!-- Header -->
      <h2 class="title">Robust 6D Object Pose Estimation in Low-Light Environments</h2>
      <img src="image/logo.png" class="logo" alt="Logo">

      <!-- Images -->
      <div class="image-row">
        <div class="image-column">
          <img src="image/objects.png" alt="20 object models overview">
          <p class="center-text">Figure 1: Our dataset covers object models with diverse challenges.</p>
        </div>
        <div class="image-column">
          <img src="image/sample.png" alt="Dark6D sample images">
          <p class="center-text">Figure 2: Samples of Dark6D dataset. Corresponding normal-light image on the left.</p>
        </div>
      </div>

      <!-- Main Content -->
      <h3>Motivation</h3>
      <p>
        6D object pose estimation is essential for applications such as robotics and augmented reality,
        where accurate spatial understanding of objects is required. Most existing methods assume
        well-lit environments and suffer performance degradation under low-light conditions due
        to weak textures, increased noise, and unreliable depth observations. These challenges lead
        to unstable feature representations and inaccurate pose predictions. Developing robust pose
        estimation approaches that remain reliable under low-light conditions is crucial for real-world
        deployment.
      </p>

      <h3>Task</h3>
      <p>
        This benchmark proposes a task of 6D pose estimation from RGB-D images in real time.
        We provide datasets containing RGB images, depth maps, segmentation masks, and ground-truth pose annotations.
        Researchers can use these datasets to explore different methods.
      </p>

      <h3>Dataset</h3>
      <p>
        The dataset includes 20 rich and low-texture objects. Images are rendered at $640 \times 480$,
        resulting in 60k samples, including low-light and normal-light images. Camera positions are uniformly
        sampled in a spatial range while pointing to the scene center for visibility. Each frame records
        RGB, depth, segmentation, and pose. This provides diverse appearance and multi-view geometric observations.
      </p>

      <a class="button button-primary button-full"
         href="https://drive.google.com/open?id=1hqU_aUEMv4dQD7hehjm3BYU0CehsiFe5"
         target="_blank">Download the datasets</a>

      <h3>Registration</h3>
      <p>
        To participate, please
        <a href="mailto:h.yuan@uu.nl?subject=SHREC 6D pose estimation: Registration" target="_blank">
          send us an email
        </a> confirming your interest and optionally your affiliation and co-authors.
      </p>

      <h3>Submission</h3>
      <p>
        Submit results before the deadline with a one-page description of your method. Results
        should be in a .txt file with image names, estimated 6D poses, and estimation speed (ms per pose).
      </p>

      <h3>Evaluation</h3>
      <p>
        We use ADD and ADD-S metrics. Given ground truth rotation R and translation T, and estimated
        rotation R_e and translation T_e, ADD computes mean distances between 3D points transformed
        by [R_e|T_e] and [R|T]. ADD-S handles symmetry-invariant pose errors.
      </p>

      <h5>Organizers</h5>
      <ul>
        <li>Honglin Yuan <sup>1</sup></li>
        <li>Yuting Zhang <sup>1</sup></li>
      </ul>
      <p>1: Nanjing University of Information Science and Technology, School of Computer Science and School of Software</p>
      <a class="button button-primary"
         href="mailto:h.yuan@uu.nl?subject=SHREC 6D pose: Question"
         target="_blank">Contact us</a>

      <!-- Footer -->

    </section>
  </div>
</body>
</html>