Ask a Question

Prefer a chat interface with context about you and your work?

A Rotation-Invariant Framework for Deep Point Cloud Analysis

A Rotation-Invariant Framework for Deep Point Cloud Analysis

Recently, many deep neural networks were designed to process 3D point clouds, but a common drawback is that rotation invariance is not ensured, leading to poor generalization to arbitrary orientations. In this paper, we introduce a new low-level purely rotation-invariant representation to replace common 3D Cartesian coordinates as the network …