3D Reconstruction Completion Using Per-Object Priors

Lead Research Organisation: Imperial College London
Department Name: Dept of Computing

Abstract

Knowledge of 3D scene geometry, including the structure of unobserved regions, is critically important to robotics, with applications in manipulator grasp planning and free space mapping. This thesis proposes novel approaches on using learning-based priors for per-object reconstructions, allowing for scalable multi-view reconstructions of entire scenes.


computer vision

Publications

10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/N509206/1 01/10/2015 30/09/2021
2287643 Studentship EP/N509206/1 01/10/2015 30/09/2019 Andrea Nicastro