MV-RoMa: From Pairwise Matching into Multi-View Track Reconstruction

Jongmin Lee, Seungyeop Kang, Sungjoo Yoo

Establishing consistent correspondences across images is essential for 3D vision tasks such as structure-from-motion (SfM), yet most existing matchers operate in a pairwise manner,…