Datasets:
VIMA
/

ArXiv:
License:
VIMA-Data / README.md
yunfanj's picture
update bibtex
f35cee8
metadata
license: cc-by-4.0

Dataset Card for VIMA-Data

Table of Contents

Dataset Description

Dataset Summary

This is the official dataset used to train general robot manipulation agents with multimodal prompts, as presented in paper. It contains 650K trajectories for 13 tasks in VIMA-Bench. All demonstrations are generated by oracles.

Dataset Structure

Data are grouped into different tasks. Within each trajectory's folder, there are two folders rgb_front and rgb_top, and three files obs.pkl, action.pkl, and trajectory.pkl. RGB frames from a certain perspective are separately stored in corresponding folder. obs.pkl includes segmentation and state of end effector. action.pkl contains oracle actions. trajectory.pkl contains meta information such as elapsed steps, task information, and object information. Users can build their custom data piepline starting from here. More details and examples can be found here.

Dataset Creation

All demonstrations are generated by scripted oracles.

Additional Information

Licensing Information

This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

Citation Information

If you find our work useful, please consider citing us!

@inproceedings{jiang2023vima,
  title     = {VIMA: General Robot Manipulation with Multimodal Prompts},
  author    = {Yunfan Jiang and Agrim Gupta and Zichen Zhang and Guanzhi Wang and Yongqiang Dou and Yanjun Chen and Li Fei-Fei and Anima Anandkumar and Yuke Zhu and Linxi Fan},
  booktitle = {Fortieth International Conference on Machine Learning},
  year      = {2023}
}