Skip to content

feat: update new #5

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
LiXiling merged 1 commit into from
May 28, 2026
Merged

feat: update new #5

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 1 addition & 14 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,19 +1,6 @@
# Project Website

This repository jekyll-ized the source code for the [Nerfies website](https://nerfies.github.io).
You only need to change the content of [index.md](/index.md).
It's possible to only write in markdown, but you can also use HTML to achieve fancier effects.
This ist the static website for github pages.

This webpage is served at [https://intuitive-robots.github.io/mops/](https://intuitive-robots.github.io/mops/)

## Test it locally

Install [Jekyll](https://jekyllrb.com/docs/installation/), and run
```
jekyll serve
```
in this directory.
Then you can see the website at `http://127.0.0.1:4000`.

## Activate Github Pages:
Go into repository settings, Github pages and serve.
97 changes: 50 additions & 47 deletions index.html
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,7 @@
<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>
<title>MOPS: Multi-Objective 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"
Expand All @@ -14,16 +14,18 @@

<!-- Open Graph / link previews -->
<meta property="og:type" content="website">
<meta property="og:title" content="MOPS: Multi-Object Photoreal Simulation Dataset">
<meta property="og:title" content="MOPS: Multi-Objective 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/">
<meta property="og:image" content="https://intuitive-robots.github.io/mops/static/image/rgb.png">

<!-- 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.">
<meta name="twitter:image" content="https://intuitive-robots.github.io/mops/static/image/rgb.png">

<!-- Fonts & icons -->
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" rel="stylesheet">
Expand Down Expand Up @@ -66,7 +68,7 @@
<div class="column has-text-centered">

<h1 class="title is-2 publication-title">
MOPS: Multi-Object Photoreal Simulation Dataset<br>
MOPS: Multi-Objective Photoreal Simulation Dataset<br>
for Computer Vision in Robot Manipulation
</h1>

Expand All @@ -92,12 +94,19 @@ <h1 class="title is-2 publication-title">
</a>
</span>
<span class="link-block">
<a href="https://github.com/LiXiling/mops-data"
<a href="https://github.com/LiXiling/mops-repo"
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>
<span class="link-block">
<a href="https://huggingface.co/collections/LiXiling/mops"
class="external-link button is-normal is-rounded is-dark">
<span class="icon"><i class="fas fa-database"></i></span>
<span>Dataset</span>
</a>
</span>
</div>
</div>

Expand All @@ -118,18 +127,23 @@ <h1 class="title is-2 publication-title">
<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>&ldquo;graspable&rdquo;</em> or <em>&ldquo;pushable&rdquo;</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.
Datasets providing high-quality visual annotations in manipulation-relevant scenes
remain scarce. We introduce <strong>MOPS</strong>, a dataset generation framework that
combines 3D assets from PartNet-Mobility and RoboCasa with a zero-shot LLM-based
augmentation pipeline to automatically normalize object scale and generate part-level
affordance annotations, describing how an object part can be manipulated (e.g., a mug
handle is <em>&ldquo;graspable,&rdquo;</em> a drawer is
<em>&ldquo;pullable&rdquo;</em>).
Built on ManiSkill3, MOPS produces photorealistic indoor scenes with pixel-perfect
ground truth for class, part, and instance segmentation, multi-label affordances, depth,
surface normals, and 6D poses, spanning 54 affordance types across 137 object
categories. Human verification confirms <strong>97.3%</strong> accuracy of the zero-shot
affordance labels. We validate MOPS on three vision benchmarks of increasing scene
complexity and show that ground-truth affordance masks improve imitation learning
success rates on 24 RoboCasa manipulation tasks by <strong>7.9 percentage
points</strong>
over RGB-only baselines, with predicted affordances still yielding measurable gains.
The dataset and framework are publicly available.
</p>
</div>
</div>
Expand Down Expand Up @@ -201,7 +215,8 @@ <h3 class="feature-title">Photorealistic Simulation</h3>
<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>
automatic part-level labeling, scale normalization, and semantic understanding
&mdash; <strong>97.3%</strong> accurate against human verification.</p>
</div>
</div>
<div class="column is-half">
Expand All @@ -218,7 +233,7 @@ <h3 class="feature-title">Multi-Modal Ground Truth</h3>
<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>
scenarios spanning 137 object categories and 54 affordance labels.</p>
</div>
</div>
</div>
Expand Down Expand Up @@ -260,7 +275,7 @@ <h2 class="section-heading">Results</h2>
</tr>
<tr class="table-section-divider table-highlight">
<td><strong>MOPS (Total)</strong></td>
<td><strong>56</strong></td>
<td><strong>54</strong></td>
<td><strong>137</strong></td>
<td>3,353</td>
</tr>
Expand Down Expand Up @@ -332,43 +347,31 @@ <h2 class="section-heading">Getting Started</h2>
<span class="tag is-warning"><strong>Alpha</strong></span>
</div>
<div class="column">
<strong>Early release</strong> &mdash; API may change. Code is split across two
repositories:
<strong>Early release</strong> &mdash; actively developed, API may change. Code and
dataset are now publicly available:
</div>
</div>
<div class="repo-links" style="margin-top:0.6rem;">
<a href="https://github.com/LiXiling/mops-data" class="repo-pill is-available">
<a href="https://github.com/LiXiling/mops-repo" class="repo-pill is-available">
<span>⚙️</span>
<span><code>mops-data</code> &mdash; Image generation in ManiSkill3</span>
<span><code>mops-repo</code> &mdash; Generation framework &amp; benchmarks</span>
<span class="repo-status">Available</span>
</a>
<a href="https://huggingface.co/collections/LiXiling/mops"
class="repo-pill is-available">
<span>🤗</span>
<span>HuggingFace &mdash; MOPS dataset collection</span>
<span class="repo-status">Available</span>
</a>
<span class="repo-pill is-coming-soon">
<span>🤖</span>
<span><code>mops-il</code> &mdash; 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 &nbsp;&middot;&nbsp;
CUDA-compatible GPU &nbsp;&middot;&nbsp; 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 &rarr;
</a>
</p>
</div>
<p class="has-text-centered">
<a href="https://github.com/LiXiling/mops-repo#installation"
class="external-link button is-link is-light is-small is-rounded">
📖 Setup &amp; Installation Guide &rarr;
</a>
</p>
</div>
</div>

Expand Down
Loading
Loading