CVPR2021

From Points to Multi-Object 3D Reconstruction

Francis Engelmann, Konstantinos Rematas, Bastian Leibe, Vittorio Ferrari

Abstract

Figure 1: We propose a single-stage model for realistic multi-object 3D reconstruction from a single RGB image. The model detects object center-points and performs reconstruction by jointly estimating 9-DoF bounding boxes and representation-agnostic 3D shape exemplars.