Abstract: This study presents a novel deep-learning framework designed to efficiently generate high-fidelity three-dimensional (3D) point clouds of rail fasteners. The proposed method overcomes limitations associated with traditional ...