GenNBV: Generalizable Next-Best-View Policy for Active 3D Reconstruction

Free-space Capturing Cross-dataset Generalization on Novel Objects z x y yaw pitc h Previous NBV Policies Our Generalizable Free-space NBV Policy Train on Hundreds of Objects Per-scene Training Object-Centric Capturing NeRF Ungeneralizable self-occlusion Figure 1. To determine the best view for 3D reconstruction, previous methods only chose from hand-crafted action space or based on object-centric capturing, lacking the ability to generalize to unforeseen scenes (Left). With our end-to-end trained, generalized free-space policy, it can generalize to unseen objects, enabling the captured drone to image from any viewpoint (Right).