Robotic grasp dataset for multi-object multi-grasp evaluation with RGB-D data.
This dataset is annotated using the same protocal as Cornell Dataset, and can be used as multi-object extension of Cornell Dataset.

Download RGB-D data and put into grasp_multiObject/rgbd/

Each testing data has one RGB image (rgb_xxxx) and one depth image (depth_xxxx).
The corresponding grasp annotation (rgb_xxxx_annotations) can be found in grasp_multiObject/annotation/

mkdir rgd
run rgbd2rgd

you will have RG-D data in grasp_multiObject/rgd/

mkdir rgd_cropped320
mkdir rgb_cropped320
run image2txt

you will have cropped RGB and RGD images in grasp_multiObject/rgd_cropped320/ and grasp_multiObject/rgb_cropped320/, respectively.

also, you will have corresponding annotation files, as well as a full list of image path.

run visualizationGripper

this file shows a simple example to visualize ground truth grasps

git clone https://github.com/ivalab/grasp_annotation_tool

you can annotate grasps on your own data with this simple tool!
Both dataset and annotation tool can also be found here

If you find it helpful for your research, please consider citing:

@inproceedings{chu2018deep,
  title = {Real-World Multiobject, Multigrasp Detection},
  author = {F. Chu and R. Xu and P. A. Vela},
  journal = {IEEE Robotics and Automation Letters},
  year = {2018},
  volume = {3},
  number = {4},
  pages = {3355-3362},
  DOI = {10.1109/LRA.2018.2852777},
  ISSN = {2377-3766},
  month = {Oct}
}

If you encounter any questions, please contact me at fujenchu[at]gatech[dot]edu