1 paper
Tianyu Chen, Wei Xiang, Kang Han +4
Feed-forward 3D reconstruction offers substantial runtime advantages over per-scene optimization, which remains slow at inference and often fragile under sparse views. However, exi…