ShapeNet

ShapeNet is a 3D object data set discussed in E244｜端到端vs上下分层：机器人路径之争，正在转向？. The source presents it as a successor to the ImageNet-style idea of large structured data, but moved from 2D images toward object shape.

The source also treats ShapeNet as insufficient for full robot manipulation by itself. Robot policies need not only shapes, but also object parts, articulation, material, friction, elasticity, and contact dynamics, which leads the episode toward PartNet / PartNet Mobility and Structured 3D Robot Data.

Connections

• Su Hao / 苏浩 - researcher context in the source.
• PartNet / PartNet Mobility - next data-set layer around parts and mobility.
• Structured 3D Robot Data and Sim2Real - why the data set matters for robot training.