kuanfang / procedural-objects

Procedural object generation for robotic manipulation

Procedural Object Generation

The procedural object generation used in Learning Task-Oriented Grasping for Tool Manipulation from Simulated Self-Supervision.

Citation

@inproceedings{fang2018rss,
author = {Kuan Fang, Yuke Zhu, Animesh Garg, Andrey Kuryenkov, Viraj Mehta, Li Fei-Fei, Silvio Savarese},
title = {Learning Task-Oriented Grasping for Tool Manipulation from Simulated Self-Supervision},
booktitle = {Robotics: Science and Systems (RSS)},
month = {June},
year = {2018}
}

Prepare

1. Normalize .obj mesh files:

   python normalize_mesh.py --input data/example/duck_vhacd.obj --output outputs/

2. Visualize a .urdf or .obj file:

   python visualize.py --input outputs/000000/body.urdf

   or

   python visualize.py --input outputs/000000/head.obj

Generation

1. Generate realistic objects:

   python generate.py --body t --color realistic --mesh data/example/duck_vhacd.obj --output outputs/ --num 1