Unified Panoramic Geometry Estimation via Multi-View Foundation Models
For computer vision researchers, PaGeR provides a unified framework that extends powerful 3D foundation models to panoramic imagery, enabling full 360-degree scene reconstruction from a single image.
PaGeR adapts a pre-trained 3D foundation model for perspective images to handle panoramic images, achieving state-of-the-art performance in single-view 360-degree geometry estimation (depth, normals, sky masks) across indoor and outdoor scenes with strong zero-shot generalization.
Geometry estimation from perspective images has greatly advanced, maturing to the point where off-the-shelf foundation models are able to reconstruct 3D scene structure not only from multi-view imagery, but even from a single view. A natural extension is 3D reconstruction from panoramas, with the exciting prospect of recovering a full 360-degree scene from a single panoramic image. In this work, we introduce PaGeR (Panoramic Geometry Reconstruction), a framework to lift powerful 3D foundation models designed for perspective imagery to the panorama domain. Our strategy is to start from a pre-trained transformer for 3D reconstruction and turn it into a unified high-performance model that predicts scale-invariant depth, metric depth, surface normals, and sky masks from both perspective and omnidirectional images, in a single forward pass. By keeping architectural changes to a minimum and mixing perspective and panoramic images during training, PaGeR retains the rich 3D prior of the underlying foundation model while learning to also estimate geometrically consistent 360-degree scenes from single panoramas. We extensively test our method in both indoor and outdoor environments and find that it delivers state-of-the-art performance and excellent zero-shot performance across a wide range of scenes.