Su Hao / 苏浩

Su Hao is introduced in E244｜端到端vs上下分层：机器人路径之争，正在转向？ as a researcher who helped connect the ImageNet data-set mindset to 3D object understanding. The episode uses his trajectory to explain why robotics needs more than internet-scale 2D image and video data.

In the source’s account, ShapeNet, PartNet / PartNet Mobility, and PartNet Mobility extend object data from 2D recognition toward 3D shape, object parts, and mobility constraints, but still fall short of the full geometry, material, friction, elasticity, and dynamics needed for Open-World Robot Manipulation.

Connections

• ImageNet - 2D data-set precedent named by the source.
• ShapeNet and PartNet / PartNet Mobility - 3D data-set lineage connected to his work in the episode.
• Structured 3D Robot Data - concept this source builds from the data-set discussion.