Skip to content

Latest commit

 

History

History
73 lines (59 loc) · 5.27 KB

README.md

File metadata and controls

73 lines (59 loc) · 5.27 KB
Raw

PoseIt Dataset

PoseIt is a visual tactile dataset of various objects with diffferent holding poses.

CC BY-SA 4.0   License: MIT       

Website • Dataset Format • Maintainers

PoseIt dataset contains RGB-D, tactile - Gelsight & WSG-DSA, force, torque, joint angle, joint velocity, and gripper force data for various grasp holding poses. In total, PoseIt consists of 1840 grasp datapoints collected from 26 distinct objects, with 16 different holding poses for each object. The PoseIt data collection pipeline contains the following phases: grasping the object, moving the object to a holding pose, and shaking the object to check its stability. To test grasp quality on a diverse range of objects, we collected data for 26 various household objects with a diverse range of size, shape, material, mass, and texture.

Dataset Format

Visualization of the data collection setup and data modalities. We use a Universal Robotics UR5e 6-DoF robot arm. An OnRobot Hex 6-Axis force/torque (F/T) sensor is attached to the end effector and records F/T measurements of the grasped objects. We use a high resolution visuo-tactile GelSight sensor. We use the Servo-electric 2-finger parallel gripper WSG50 from Weiss Robotics to attach these tactile sensors. We utilize two Azure Kinects and one RGB camera to visually capture the robot's workspace.

Maintainers