Global Structure-from-Motion Meets Feedforward Reconstruction

Researchers propose a hybrid Structure-from-Motion pipeline that combines classical methods with feedforward 3D reconstruction.

The paper compares the strengths and weaknesses of both approaches and shows how they can complement each other.

Experiments report strong gains and state-of-the-art results on multiple benchmarks for camera pose and scene recovery.

The work targets long-standing failure cases in 3D reconstruction while improving robustness, scalability, and overall performance.