arXiv 2510.07856
XYZCylinder: Feedforward Reconstruction for Driving Scenes Based on A Unified Cylinder Lifting Method
By Haochen Yu, Qiankun Liu, et al.
Published 2025-10-09
Wiki summary
Explore the paper's summary, context, and related research on Papiers.
Recently, more attention has been paid to feedforward reconstruction paradigms, which mainly learn a fixed view transformation implicitly and reconstruct the scene with a single representation. However, their generalization capability and reconstruction accuracy are still limited while reconstructing driving scenes, which results from two aspects: (1) The fixed view transformation fails when the camera configuration…