Point Set Voting for Partial Point Cloud Analysis

The continual improvement of 3D sensors has driven the development algorithms to perform point cloud analysis. In fact, techniques for classification and segmentation have in recent years achieved incredible performance part by leveraging large synthetic datasets. Unfortunately these same state-of-the-art approaches poorly when applied incomplete clouds. This limitation existing is particularly concerning since clouds generated real world are usually due perspective view or occlusion other objects. paper proposes a general model partial analysis wherein latent feature encoding complete inferred applying set voting strategy. particular, each local constructs vote that corresponds distribution space, optimal one with highest probability. approach ensures any subsequent robust observation while simultaneously guaranteeing proposed able output multiple possible results. illustrates this method achieves on shape classification, completion.