Robust and Efficient Structure-from-Motion Method for Ambiguous Large-Scale Indoor Scene

摘要

Figure 1: Comparison of reconstruction results of large-scale indoor scene with high ambiguity and repeated structures. (a) Result of COLMAP [Schönberger and Frahm 2016], (b) result of [Su et al. 2020], and (c) result of our method. Both (a) and (b) indicate failure to reconstruct some ambiguous areas and floor collapse. By contrast, our method could successfully construct a complete model, and all parts were connected correctly. ACM Reference Format: Chia-Hung Yu and Kuan-Wen Chen. 2022. Robust and Efficient Structurefrom-Motion Method for Ambiguous Large-Scale Indoor Scene. In SIGGRAPH Asia 2022 Posters (SA ’22 Posters), December 06-09, 2022. ACM, New York, NY, USA, 2 pages. https://doi.org/10.1145/3550082.3564199