Object Models

The models of the objects are available at ycb-benchmarks.s3-website-us-east-1.amazonaws.com.

In order to ease adoption, we collect visual data that is commonly required for grasping algorithms and generate 3D models for use in simulation. We use the scanning rig used to collect the BigBIRD dataset [1]. The rig, shown in the figure below, has 5 RGBD sensors and 5 high-resolution RGB cameras arranged in a quarter-circular arc.

rig

Figure 1: BigBIRD Object Scanning Rig: the box contains a computer-controlled turntable.

 

Each object is placed on a computer-controlled turntable, which is rotated by 3 degrees at a time, yielding 120 turntable orientations. Together, this yields 600 RGBD images and 600 high-resolution RGB images. The process is completely automated, and the total collection time for each object is under 5 minutes.

In total, for each object, we provide:

  • 600 RGBD images
  • 600 high-resolution RGB images
  • Segmentation masks for each image
  • Calibration information for each image
  • Texture-mapped 3D mesh models

 

models

Figure 2: Point cloud and textural data overlays on two YCB objects: mustard bottle and power drill.

 

[1] A. Singh, J. Sha, K. S. Narayan, T. Achim, and P. Abbeel, “BigBIRD: A Large-Scale 3D Database of Object Instances,” International Conference on Robotics and Automation, 2014.