Holo360D: A Large-Scale Real-World Dataset with Continuous Trajectories for Advancing Panoramic 3D Reconstruction and Beyond

Holo360D provides continuous panoramic trajectories for studying feed-forward 3D reconstruction. The dataset contains 109,495 panoramas paired with point clouds, meshes and camera poses. Scanner and panoramic-camera measurements are aligned through SLAM, followed by geometry cleaning and mesh refinement. Reconstruction experiments investigate how fine-tuning on these sequences helps models handle panoramic imagery and its spherical distortions.