Point Cloud Completion
Overview
Point cloud completion estimates unseen surfaces or objects from incomplete 3D scans, supporting tasks such as 3D perception, scene mapping, and anatomical reconstruction. Methods range from coarse-to-fine networks and Transformer or state-space models to diffusion and flow-based generators, often using shape priors, temporal observations, or images to guide reconstruction. Object benchmarks such as PCN and ShapeNet test shape recovery, while datasets such as SemanticKITTI and KITTI-360 target large-scale LiDAR scene completion. A central challenge is balancing plausible recovery of missing geometry with fidelity to observed points, especially under real-world noise and occlusion, while keeping inference efficient.
Latest papers
30 papers4 papers in the last four weeks, up 33% on the four weeks before. 0.0% of all new papers.