FastMap: Revisiting Structure from Motion through First-Order Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Jiahao, Wang, Haochen, Irshad, Muhammad Zubair, Vasiljevic, Igor, Walter, Matthew R., Guizilini, Vitor Campagnolo, Shakhnarovich, Greg
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911112930263040
author Li, Jiahao
Wang, Haochen
Irshad, Muhammad Zubair
Vasiljevic, Igor
Walter, Matthew R.
Guizilini, Vitor Campagnolo
Shakhnarovich, Greg
author_facet Li, Jiahao
Wang, Haochen
Irshad, Muhammad Zubair
Vasiljevic, Igor
Walter, Matthew R.
Guizilini, Vitor Campagnolo
Shakhnarovich, Greg
contents We propose FastMap, a new global structure from motion method focused on speed and simplicity. Previous methods like COLMAP and GLOMAP are able to estimate high-precision camera poses, but suffer from poor scalability when the number of matched keypoint pairs becomes large, mainly due to the time-consuming process of second-order Gauss-Newton optimization. Instead, we design our method solely based on first-order optimizers. To obtain maximal speedup, we identify and eliminate two key performance bottlenecks: computational complexity and the kernel implementation of each optimization step. Through extensive experiments, we show that FastMap is up to 10 times faster than COLMAP and GLOMAP with GPU acceleration and achieves comparable pose accuracy.
format Preprint
id arxiv_https___arxiv_org_abs_2505_04612
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle FastMap: Revisiting Structure from Motion through First-Order Optimization
Li, Jiahao
Wang, Haochen
Irshad, Muhammad Zubair
Vasiljevic, Igor
Walter, Matthew R.
Guizilini, Vitor Campagnolo
Shakhnarovich, Greg
Computer Vision and Pattern Recognition
We propose FastMap, a new global structure from motion method focused on speed and simplicity. Previous methods like COLMAP and GLOMAP are able to estimate high-precision camera poses, but suffer from poor scalability when the number of matched keypoint pairs becomes large, mainly due to the time-consuming process of second-order Gauss-Newton optimization. Instead, we design our method solely based on first-order optimizers. To obtain maximal speedup, we identify and eliminate two key performance bottlenecks: computational complexity and the kernel implementation of each optimization step. Through extensive experiments, we show that FastMap is up to 10 times faster than COLMAP and GLOMAP with GPU acceleration and achieves comparable pose accuracy.
title FastMap: Revisiting Structure from Motion through First-Order Optimization
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2505.04612