YCB dataset with urdf files. Ready to be used in different simulations such as PyBullet.
Have ycb folder in your project directory.
For convenience, the locations of urdf file can be found with __init__.py of ycb directly. An example is
from ycb import YCB
cracker_box_urdf = YCB["cracker_box"]
You need to have pathlib
installed.
pip install pathlib
Here is the procedure of how this repository is made.
- The dataset is downloaded from the YCB dataset website with script
download_ycb_dataset.py
provided by ycb-tools. Only textured models with Google 16K are used. - The urdf files are generated with scripts
build_object_library.py
andobject_urdf.py
provided by object2urdf. - The mass of each object is manually set based on the provided mass form. The inertia of all objects are simply diagonal matrix with value 1e-3.
- All the texture images are compressed to 80% JPG with image size unchanged with
image_compressor.py
. This should improve simulation performance and save space without affecting the vision pipeline.
This repository only includes models with Google 16K. The other models, shown as below, are not included.
- models only with rgbd, not google_16k:
- 001
- 041
- 049
- 072-f
- 072-h
- 072-i
- 072-j
- 072-k
- 073-h
- 073-i
- 073-j
- 073-k
- 073-l
- 073-m
- low-quality models:
- 023
- 039
- 046
- 047
- 063-c
- 063-d
- 063-e
- 063-f
- 072-g