GLArt: Grasp Labels for ARTiculated objects

Repository is part of CenterArt paper, for generating a dataset of robust object-centric 6-DoF grasp poses for articulated objects. The dataset consists of 82 articulated objects collected from the PartNet Mobility dataset, spanning five different categories: Microwave, Oven, Refrigerator, Dishwasher, and Storage Furniture. Overall, 375,266 grasp labels are generated for 766 object-joint state pairs.

Dataset Download

Download the modified urdfs and generated grasp labels:
urdfs
grasps

Installation

If you want to generate the grasp labels for new objects, first install the required libraries:

conda create --name glart_env python=3.8
conda activate glart_env
git clone [email protected]:PRBonn/manifold_python.git
cd manifold_python
git submodule update --init
make install
cd GLArt
pip install -r requirements.txt

Dataset Generation

First, put the urdfs in datasets/urdfs and update the configs/objects_info.json file. Then, generate and pointclouds of articulated points:

python pointcloud_generator.py

Finally, generate object-centric 6-DoF grasp labels:

python grasp_generator.py

Citation

@inproceedings{mokhtar2024centerart,
  title={CenterArt: Joint Shape Reconstruction and 6-DoF Grasp Estimation of Articulated Objects},
  author={Mokhtar, Sassan and Chisari, Eugenio and Heppert, Nick and Valada, Abhinav},
  booktitle={ICRA 2024 Workshop on 3D Visual Representations for Robot Manipulation}
}

Feedback

For any feedback or inquiries, please contact [email protected]