[robotics-worldwide] [news] Large-Scale Robotic Grasping Database is out!
jeannette.bohg at tuebingen.mpg.de
Wed Aug 26 11:03:28 PDT 2015
Dear fellow Roboticists,
we would like to draw your attention to the release of a new large-scale robotic grasping database at http://grasp-database.dkappler.de.
It provides grasps that are applied to more than 700 distinct objects from over 80 categories. These grasps are generated in simulation and evaluated using the standard epsilon-metric and a new physics-metric. In crowdsourcing experiments, we have confirmed that the proposed physics-metric is a more consistent predictor for grasp success than the epsilon-metric.
In total, the database provides around 500.000 labeled grasp each annotated with stability labels from these different metrics.
Additionally, we simulate noisy and incomplete perception of objects from different viewpoints using a realistic model of an RGB-D camera.
This allows us to additionally link representations of local object shape to each grasp.
This database provides a very interesting dataset for learning how to grasp with techniques that can leverage big data.
For more detailed information, please refer to .
The database is developed and provided to you by the Autonomous Motion Department of the Max Planck Institute for Intelligent Systems.
It is freely available at http://grasp-database.dkappler.de.
* Docker Container for easy installation of all dependencies
* Python Interface
* Data Visualization tool
* Efficient HDF5 database format
* Catkin Workspace
* Extendibility (Objects, Hands, Feature Representations)
Best regards, Daniel Kappler and Jeannette Bohg.
 Daniel Kappler, Jeannette Bohg and Stefan Schaal. Leveraging Big Data for Grasp Planning.
In: Proceedings of the IEEE International Conference on Robotics and Automation. 2015. Seattle, WA, USA.
More information about the robotics-worldwide