2026
Selfi: Self-improving Reconstruction Engine via 3D Geometric Feature Alignment
CVPR 2026
Novel View Synthesis (NVS) has traditionally relied on models with explicit 3D inductive biases combined with known camera parameters from Structure-from-Motion (SfM) beforehand. Recent vision foundation models like VGGT take an orthogonal approach -- 3D knowledge is gained implicitly through traini