3D Shape Modeling and Analysis

Computational approaches to 3D shape modeling and analysis are concerned with how computers represent, reconstruct, and reason about the geometry of physical objects — whether captured as raw point clouds from a sensor, encoded as polygon meshes, or implicitly learned inside a neural network. Getting this right matters for robotics, medical imaging, autonomous driving, and digital fabrication, where accurate geometric understanding directly affects what a system can safely do in the world. Recent work has converged on deep learning architectures that operate natively on irregular 3D data, with neural radiance fields and implicit surface representations pushing reconstruction quality well beyond what classical mesh pipelines could achieve. Open challenges include making these models generalize across object categories with limited supervision, and closing the gap between the clean reconstructions produced in controlled settings and the noisy, incomplete data encountered in real deployments.

Works

47,462

Total citations

613,630

Keywords

Deep LearningPoint Clouds3D ReconstructionMesh SegmentationShape RepresentationNeural Radiance Fields