LabelArticulation is an interactive articulated object modeling tool that implements:
- Align origin and axis of raw scanned point cloud
- Interactive part segmentation
- Joint annotation (joint location and direction)
- URDF file export
We also support rigid and articulated object 6D pose labeling from captured RGB-D images
$ pip install -r requirements.txt
- 4/2021: Add rigid and articulated object 6D pose labeling tools
LabelArticulation is released under the MIT license.
If you use LabelArticulation in your research to build dataset, please use the following BibTeX entry.
@misc{LabelArticulation2021,
author = {Liu Liu and Han Xue and Wenqiang Xu},
title = {LabelArticulation},
howpublished = {\url{https://github.com/liuliu66/LabelArticulation}},
year = {2021}
}
After installation, please see Tutorial (articulation modeling)
for brief tutorials on how to model the raw scanned point cloud into articulation.
please see Tutorial (articulation pose label)
for brief tutorials on how to label 6D pose of articulated objects.
We strongly recommend to follow pyrs repository. Images of directory "data/" were captured using that repository.
- output a total transformation matrix
- add input arguments (model name, scene name)
- add input argument of initial pose
- add part-level 6D pose annotation
- add prismatic and revolute joint annotation
- add function to choose various models
- visualize depth data of input scene
- visualize coordinate axis
- realsense to capture rgbd images
- annotate point cloud and mesh at the same time