-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
418 lines (377 loc) · 21.8 KB
/
index.html
File metadata and controls
418 lines (377 loc) · 21.8 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
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
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>MOPS: Multi-Object Photoreal Simulation Dataset for Computer Vision in Robot Manipulation</title>
<meta name="description"
content="MOPS is a photorealistic simulation dataset providing comprehensive ground truth annotations — RGB, depth, normals, part segmentation, and affordance labels — for robot manipulation research.">
<meta name="keywords"
content="robot manipulation, dataset, affordance, segmentation, simulation, computer vision, imitation learning, ManiSkill">
<meta name="author" content="Maximilian X. Li, Paul Mattes, Nils Blank, Rudolf Lioutikov">
<!-- Open Graph / link previews -->
<meta property="og:type" content="website">
<meta property="og:title" content="MOPS: Multi-Object Photoreal Simulation Dataset">
<meta property="og:description"
content="Photorealistic simulation dataset with part-level affordance annotations for robot manipulation.">
<meta property="og:url" content="https://intuitive-robots.github.io/mops/">
<!-- Twitter card -->
<meta name="twitter:card" content="summary_large_image">
<meta name="twitter:title" content="MOPS Dataset">
<meta name="twitter:description"
content="Photorealistic simulation dataset with part-level affordance annotations for robot manipulation.">
<!-- Fonts & icons -->
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" rel="stylesheet">
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.5.0/css/all.min.css">
<link rel="icon" type="image/svg+xml" href="./static/image/IRL_Logo_Icon_blue.svg">
<!-- Bulma CSS -->
<link rel="stylesheet" href="./static/css/bulma.min.css">
<!-- highlight.js for code blocks -->
<link rel="stylesheet"
href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/11.9.0/styles/atom-one-dark.min.css">
<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/11.9.0/highlight.min.js"></script>
<script>hljs.highlightAll();</script>
<!-- Site styles -->
<link rel="stylesheet" href="./static/css/index.css">
<!-- MathJax -->
<script>
MathJax = {
tex: {
inlineMath: [['$', '$'], ['\\(', '\\)']],
displayMath: [['$$', '$$'], ['\\[', '\\]']],
processEscapes: true,
},
svg: { fontCache: 'global' }
};
</script>
<script id="MathJax-script" async src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-chtml.js"></script>
</head>
<body>
<!-- ── Hero / Header ───────────────────────────────── -->
<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-2 publication-title">
MOPS: Multi-Object Photoreal Simulation Dataset<br>
for Computer Vision in Robot Manipulation
</h1>
<div class="is-size-5 publication-authors">
<span class="author-block">Maximilian X. Li,</span>
<span class="author-block">Paul Mattes,</span>
<span class="author-block">Nils Blank,</span>
<span class="author-block">Rudolf Lioutikov</span>
</div>
<div class="is-size-6 publication-authors mt-2">
<span class="author-block">Intuitive Robots Lab, Karlsruhe Institute of Technology,
Germany</span>
</div>
<div class="column has-text-centered mt-4">
<div class="publication-links">
<span class="link-block">
<a href="./static/Li2026_MOPS.pdf"
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>
<span class="link-block">
<a href="https://github.com/LiXiling/mops-data"
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>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- ── Main content ────────────────────────────────── -->
<section class="section">
<div class="container is-max-desktop">
<div class="content">
<!-- ① Abstract -->
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="section-heading">Abstract</h2>
<div class="has-text-justified abstract-text">
<p>
Datasets bridging computer vision and robotics by providing high-quality visual
annotations in manipulation-relevant scenes remain limited.
This work introduces the <strong>Multi-Object Photoreal Simulation (MOPS)</strong>
dataset, which provides comprehensive ground truth annotations for photorealistic
simulated environments. MOPS employs a zero-shot asset augmentation pipeline based on
Large Language Models (LLM) to automatically normalize 3D object scale and generate
part-level affordances. The dataset features pixel-level segmentations for tasks
crucial to robotic perception, including fine-grained part segmentation and affordance
prediction (e.g., <em>“graspable”</em> or <em>“pushable”</em>).
By combining detailed annotations with photorealistic simulation, MOPS generates a
vast, diverse collection of scenes to accelerate progress in robot perception and
manipulation. We validate MOPS through vision and robot learning benchmarks.
</p>
</div>
</div>
</div>
<hr class="section-divider">
<!-- ② Annotation Modalities -->
<div class="has-text-centered" style="margin-bottom:2rem;">
<h2 class="section-heading">Annotation Modalities</h2>
<p class="section-subheading">Rich, multi-modal ground truth for every scene</p>
</div>
<div class="columns is-multiline modality-gallery">
<div class="column is-half-tablet is-one-quarter-desktop">
<div class="modality-card">
<figure class="image"><img src="./static/image/rgb.png" alt="RGB render"></figure>
<div class="modality-label">
<span class="modality-dot" style="background:#4a90d9;"></span>RGB
</div>
</div>
</div>
<div class="column is-half-tablet is-one-quarter-desktop">
<div class="modality-card">
<figure class="image"><img src="./static/image/depth.png" alt="Depth map"></figure>
<div class="modality-label">
<span class="modality-dot" style="background:#e67e22;"></span>Depth
</div>
</div>
</div>
<div class="column is-half-tablet is-one-quarter-desktop">
<div class="modality-card">
<figure class="image"><img src="./static/image/normal.png" alt="Surface normals"></figure>
<div class="modality-label">
<span class="modality-dot" style="background:#27ae60;"></span>Surface Normals
</div>
</div>
</div>
<div class="column is-half-tablet is-one-quarter-desktop">
<div class="modality-card">
<figure class="image"><img src="./static/image/segm.png"
alt="Part / Affordance segmentation"></figure>
<div class="modality-label">
<span class="modality-dot" style="background:#8e44ad;"></span>Segmentation
</div>
</div>
</div>
</div>
<hr class="section-divider">
<!-- ③ Key Features + Technical Overview (merged) -->
<div class="has-text-centered" style="margin-bottom:2rem;">
<h2 class="section-heading">Key Features</h2>
</div>
<div class="columns is-multiline feature-grid" style="margin-bottom:1.5rem;">
<div class="column is-half">
<div class="feature-card">
<div class="feature-icon">🎨</div>
<h3 class="feature-title">Photorealistic Simulation</h3>
<p class="feature-text">High-quality rendering via ManiSkill3 & SAPIEN, built on a
normalized asset pipeline with automatic part-level annotation across multiple 3D
libraries.</p>
</div>
</div>
<div class="column is-half">
<div class="feature-card">
<div class="feature-icon">🤖</div>
<h3 class="feature-title">LLM-Powered Annotation</h3>
<p class="feature-text">Zero-shot asset augmentation using large language models for
automatic part-level labeling, scale normalization, and semantic understanding.</p>
</div>
</div>
<div class="column is-half">
<div class="feature-card">
<div class="feature-icon">🏷️</div>
<h3 class="feature-title">Multi-Modal Ground Truth</h3>
<p class="feature-text">RGB, depth, surface normals, part segmentation, affordance maps
(<em>graspable</em>, <em>pushable</em>, …), and 6D pose — all
pixel-aligned.</p>
</div>
</div>
<div class="column is-half">
<div class="feature-card">
<div class="feature-icon">🏠</div>
<h3 class="feature-title">Diverse Environments</h3>
<p class="feature-text">Kitchen environments, cluttered tabletops, and isolated object
scenarios spanning 137 object categories and 56 affordance labels.</p>
</div>
</div>
</div>
<hr class="section-divider">
<!-- ④ Results — dataset comparison + robot results side by side -->
<div class="has-text-centered" style="margin-bottom:2rem;">
<h2 class="section-heading">Results</h2>
</div>
<div class="columns is-multiline results-split">
<!-- Left: dataset comparison -->
<div class="column is-half">
<p class="results-col-label">Dataset Comparison</p>
<p class="results-col-sub">Taxonomic coverage vs. existing affordance datasets</p>
<div class="comparison-table-wrap">
<table class="comparison-table">
<thead>
<tr>
<th>Dataset</th>
<th>Aff.</th>
<th>Cat.</th>
<th>Obj.</th>
</tr>
</thead>
<tbody>
<tr>
<td>RGB-D Part</td>
<td>7</td>
<td>17</td>
<td>105</td>
</tr>
<tr>
<td>3D-AffNet</td>
<td>16</td>
<td>23</td>
<td><strong>22,949</strong></td>
</tr>
<tr class="table-section-divider table-highlight">
<td><strong>MOPS (Total)</strong></td>
<td><strong>56</strong></td>
<td><strong>137</strong></td>
<td>3,353</td>
</tr>
</tbody>
</table>
<p class="table-caption">MOPS leads on affordance label and category breadth; 3D-AffNet has
more raw instances.</p>
</div>
</div>
<!-- Right: robot results -->
<div class="column is-half">
<p class="results-col-label">Robot Manipulation</p>
<p class="results-col-sub">Imitation learning on 24 RoboCasa tasks · 10 seeds each</p>
<div class="results-stats-inline">
<div class="stat-box">
<div class="stat-number">21.25<span class="stat-unit">%</span></div>
<div class="stat-label">Success Rate</div>
<div class="stat-sublabel">RGB + MOPS Affordances</div>
</div>
<div class="stat-box stat-box-gain">
<div class="stat-number">+7.92<span class="stat-unit">pp</span></div>
<div class="stat-label">Absolute Gain</div>
<div class="stat-sublabel">over RGB-only baseline</div>
</div>
</div>
<div class="comparison-table-wrap" style="margin-top:1rem;">
<table class="comparison-table">
<thead>
<tr>
<th>Policy Inputs</th>
<th>Success</th>
<th>Gain</th>
</tr>
</thead>
<tbody>
<tr>
<td>RGB only</td>
<td>13.33%</td>
<td><span class="gain-neutral">—</span></td>
</tr>
<tr class="table-highlight">
<td><strong>+ MOPS Affordances</strong></td>
<td><strong>21.25%</strong></td>
<td><span class="gain-positive">+7.92</span></td>
</tr>
</tbody>
</table>
<p class="table-caption">MOPS affordances provide a consistent boost across all 24 tasks.
</p>
</div>
</div>
</div>
<hr class="section-divider">
<!-- ⑤ Getting Started (with alpha notice inline) -->
<div class="has-text-centered" style="margin-bottom:2rem;">
<h2 class="section-heading">Getting Started</h2>
</div>
<div class="columns is-centered">
<div class="column is-four-fifths">
<!-- Alpha notice — contextually placed here -->
<div class="notification is-warning is-light alpha-notice" style="margin-bottom:1.5rem;">
<div class="columns is-vcentered is-mobile">
<div class="column is-narrow">
<span class="tag is-warning"><strong>Alpha</strong></span>
</div>
<div class="column">
<strong>Early release</strong> — API may change. Code is split across two
repositories:
</div>
</div>
<div class="repo-links" style="margin-top:0.6rem;">
<a href="https://github.com/LiXiling/mops-data" class="repo-pill is-available">
<span>⚙️</span>
<span><code>mops-data</code> — Image generation in ManiSkill3</span>
<span class="repo-status">Available</span>
</a>
<span class="repo-pill is-coming-soon">
<span>🤖</span>
<span><code>mops-il</code> — Robot trajectories in RoboCasa v0.1</span>
<span class="repo-status">Coming Soon</span>
</span>
</div>
</div>
<div class="install-box">
<p class="install-prereqs">
<strong>Prerequisites:</strong> Python 3.10 ·
CUDA-compatible GPU · 16 GB+ RAM
</p>
<pre><code class="language-bash">conda create -n mops python=3.10
conda activate mops
pip install mani_skill
git clone https://github.com/LiXiling/mops-data
cd mops-data
pip install -e .</code></pre>
<p style="margin-top:1rem;">
<a href="https://github.com/LiXiling/mops-data#installation"
class="external-link button is-link is-light is-small is-rounded">
📖 Full Installation Guide →
</a>
</p>
</div>
</div>
</div>
<hr class="section-divider">
<!-- ⑥ Citation -->
<div class="has-text-centered" style="margin-bottom:2rem;">
<h2 class="section-heading">Citation</h2>
<p class="section-subheading">If you use MOPS in your research, please cite our work</p>
</div>
<div class="columns is-centered">
<div class="column is-four-fifths">
<div class="citation-box">
<pre><code class="language-bibtex">@article{li2026mops,
title = {Multi-Objective Photoreal Simulation (MOPS) Dataset
for Computer Vision in Robot Manipulation},
author = {Maximilian Xiling Li and Paul Mattes and
Nils Blank and Rudolf Lioutikov},
year = {2026}
}</code></pre>
</div>
</div>
</div>
</div><!-- /.content -->
</div><!-- /.container -->
</section>
<!-- ── Footer ──────────────────────────────────────── -->
<footer class="footer">
<div class="container">
<div class="columns is-centered">
<div class="column is-8 has-text-centered">
<p>
© 2026 Intuitive Robots Lab, Karlsruhe Institute of Technology.<br>
Dataset and code released under <a href="https://creativecommons.org/licenses/by/4.0/"
target="_blank">CC-BY 4.0</a>.
</p>
</div>
</div>
</div>
</footer>
</body>
</html>