3D Shape Completion
Overview
3D shape completion infers the unseen parts of an object or scene from incomplete observations, helping robots, 3D modeling systems, and medical tools reason about full geometry. Methods range from point-cloud, voxel, and implicit-shape models to statistical shape models and generative diffusion or flow priors, often guided by images, semantic features, or sensor physics. Evaluations span object benchmarks such as PCN and ShapeNet, indoor and driving scenes, and anatomical reconstruction from CT or ultrasound. A central challenge is balancing plausible missing geometry with faithful preservation of observed regions, especially when real-world occlusions differ from training data.
Latest papers
23 papers1 paper in the last four weeks, down 67% on the four weeks before. 0.0% of all new papers.