---

Empart: Interactive Mesh Simplification

This research project offers a web-based interactive tool for simplifying your meshes into convex shapes with varying error tolerances. The project is particularly aimed at finetuning collision geometry for physics simulation and robotics algorithms.

Read our pre-print paper explaining Empart on arXiv.

---

Getting Started

Prerequisites

• Ubuntu

Ensure you have these installed:

• Node.js: Download from the official Node.js website or

  sudo apt install nodejs npm

• Conda (Anaconda or Miniconda): Download Miniconda from the Conda website. or

  curl -O https://repo.anaconda.com/archive/Anaconda3-2025.06-0-Linux-x86_64.sh
  bash ~/Anaconda3-2025.06-0-Linux-x86_64.sh

Installation

1. Clone the Repository:

   git clone https://github.com/intrinsic-opensource/empart.git
   cd empart

2. Set Up Python Environment with Conda:

   conda create -n empart python=3.11
   conda activate empart
   pip install -r requirements.txt

3. Install Node.js Dependencies:

   cd tools/interactive-viewer
   npm install

4. Install Blender (optional for watertight mesh inputs):

   Note: If you are expecting to simplify meshes that are not watertight*, our tool requires blender to compute mesh intersections. Install blender from https://www.blender.org/download/

   tar -xf blender-4.5.1-linux-x64.tar.xz
   export PATH="/home/youruser/Downloads/blender-4.5.1-linux-x64:$PATH"
   source ~/.bashrc
   

   *A watertight triangle mesh is a closed 2-manifold or equivalently, every edge is shared by two faces.

---

Running the Application

cd tools/interactive-viewer
npm start

This command will start a local development server and open the application in your browser.

Running Tests

Install playwright which is used to replay mouse and web events.

playwright install

Then, run the tests using pytest:

pytest

---

Citing Empart

Please include this bibtex entry when citing Empart:

@misc{vu2025empartinteractiveconvexdecomposition,
      title={Empart: Interactive Convex Decomposition for Converting Meshes to Parts}, 
      author={Brandon Vu and Shameek Ganguly and Pushkar Joshi},
      year={2025},
      eprint={2509.22847},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2509.22847}, 
}

---

Acknowledgements

empart's ability to efficiently compute mesh intersections is made possible by amazing open source projects such as Manifold3d and Blender