Map-free Visual Relocalization: Metric Pose Relative to a Single Image
Abstract
Dataset
We make available a dataset of 655 places, collected by non-expert users world-wide. The full dataset is publicly available here.
Benchmark
We propose new metrics based on the reprojection error of virtual objects. To successfully compete in our benchmark, methods need to be precise but also provide a robust confidence estimate of their predictions. You can submit to our public leaderboard here.
BibTeX
Please cite our paper if you use our dataset or code.
@inproceedings{arnold2022mapfree,
title={Map-free Visual Relocalization: Metric Pose Relative to a Single Image},
author={Arnold, Eduardo and Wynn, Jamie and Vicente, Sara and Garcia-Hernando, Guillermo and Monszpart, {\'{A}}ron and Prisacariu, Victor Adrian and Turmukhambetov, Daniyar and Brachmann, Eric},
booktitle={ECCV},
year={2022},
}