BimanGrasp-Dataset

KO | EN

gitlite — search

#python #typescript #golang #android #machine-learning #rust #deep-learning #macos #java #windows #awesome-list #pytorch #openai #mcp

BimanGrasp-Dataset

(★ 24)

This is the official repo for the release of BimanGrasp Dataset of the RA-L and ICRA 25' Paper "Bimanual Grasp Synthesis for Dexterous Robot Hands"

BimanGrasp-Dataset 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# BimanGrasp-Dataset
This is the official repository for the BimanGrasp-Dataset release of paper

<p align="center">
  <h2 align="center">Bimanual Grasp Synthesis for Dexterous Robot Hands</h2>

<p align="center">
    <strong>Yanming Shao</strong></a>
    ·
    <strong>Chenxi Xiao*</strong>
 </p>
 
<h3 align="center">RA-L | ICRA 25'</h3>

BimanGrasp-Dataset contains diverse bimanual dexterous grasping for various objects, as shown in the following gif figures (of 6 objects)

## Introduction

BimanGrasp-Dataset is a large-scale synthetic dataset of a pair of shadow robot hands grasping various objects. All the grasps are verified with Isaac Gym simulator, and through penetration test (less than 1.5 mm). In this repo, we provide the grasp pose data together with object meshes and other assets. All the grasps can be visualized with plotly package.

## Installation

# We suggest to use conda/mamba environments for loading and visualizing bimanual grasp poses, e.g.,
    conda create -n bimangrasp python=3.7
    conda activate bimangrasp

# for cuda == 11.3
    conda install pytorch==1.12.1 torchvision==0.13.1 cudatoolkit=11.3 -c pytorch
    # else, just install the torch version that matches the cuda version
    
    # install some necessary packages
    pip install trimesh plotly numpy argparse transforms3d

# install pytorch3d that matches torch and python version, e.g., with pytorch==1.12.1:
    cd third_party && wget https://github.com/facebookresearch/pytorch3d/archive/refs/tags/v0.7.1.tar.gz && tar -xzvf v0.7.1.tar.gz && cd v0.7.1.tar.gz && pip install -e. && cd ../..

# install other third party packages
    cd third_party/pytorch_kinematics && pip install -e .
    cd ../third_party/torchSDF && pip install -e . && cd ../..

## Visualization

First, download the release dataset and object meshes from https://github.com/Tsunami-kun/BimanGrasp-Dataset/releases, and then extract them to the root directory of the repo.

# This outputs 3d html visualization. ckeck dataset for object names (<object_name>.npy), and use --object_name <object_name> --num <num> to visualize the <num>th pose for the object named by <object_name>.
    
    python visualization.py --object_name <object_name> --num <num>

One visualization example (both 2D screenshot and 3D html vis.) is in the directory examples/.

## Examples

[3D Visualization Example](examples/example.html)

![2D Screenshot Example](https://raw.githubusercontent.com/Tsunami-kun/BimanGrasp-Dataset/main/examples/example.png)

## Acknowledgement

We would like to express our gratitude to the authors of the following repositories, from which we referenced code:

* [DexGraspNet](https://github.com/PKU-EPIC/DexGraspNet/tree/main)
* [UGG](https://github.com/Jiaxin-Lu/ugg/tree/main)

## Grasp Generation Repo

Our released BimanGrasp-Generation is in this repo: [[BimanGrasp-Dataset](https://github.com/Tsunami-kun/BimanGrasp-Dataset)].

## Citation
If you find BimanGrasp-Dataset useful in your research, please cite
```
@article{shao2024bimanual,
  title={Bimanual Grasp Synthesis for Dexterous Robot Hands},
  author={Shao, Yanming and Xiao, Chenxi},
  journal={IEEE Robotics and Automation Letters},
  year={2024},
  volume={9},
  number={12},
  pages={11377-11384},
  publisher={IEEE},
  doi={10.1109/LRA.2024.3490393}
}
```

BimanGrasp-Dataset

BimanGrasp-Dataset 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

BimanGrasp-Dataset

BimanGrasp-Dataset 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README