SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2502.17237 · 2025

MegaLoc: One Retrieval to Place Them All

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 6 functions out of this paper's own repositories and ran 2 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
gmberton/VPR-methods-evaluation canonical 1 of 4
gmberton/megaloc canonical 1 of 2
FunctionStatusWhere it lives
denormalize Ran gmberton/VPR-methods-evaluation/vpr_models/utils.py
code served (permissive licence) · get_code("10c34609af1437f3")
log_otp_solver Ran gmberton/megaloc/megaloc_model.py
code served (permissive licence) · get_code("cc0d6b2496e86c33")
get_boq Not yet run gmberton/VPR-methods-evaluation/vpr_models/boq.py
code served (permissive licence) · get_code("15f6c692126a38d8")
get_dino_mix Not yet run gmberton/VPR-methods-evaluation/vpr_models/dinomix.py
code served (permissive licence) · get_code("846fe56e1c06a617")
get_matching_probs Not yet run gmberton/megaloc/megaloc_model.py
code served (permissive licence) · get_code("6d34ff3423316bc7")
get_qaa Not yet run gmberton/VPR-methods-evaluation/vpr_models/qaa.py
code served (permissive licence) · get_code("bd8f3a7438258b2b")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Retrieving images from the same location as a given query is an important component of multiple computer vision tasks, like Visual Place Recognition, Landmark Retrieval, Visual Localization, 3D reconstruction, and SLAM. However, existing solutions are built to specifically work for one of these tasks, and are known to fail when the requirements slightly change or when they meet out-of-distribution data. In this paper we combine a variety of existing methods, training techniques, and datasets to train a retrieval model, called MegaLoc, that is performant on multiple tasks. We find that MegaLoc (1) achieves state of the art on a large number of Visual Place Recognition datasets, (2) impressive results on common Landmark Retrieval datasets, and (3) sets a new state of the art for Visual Localization on the LaMAR datasets, where we only changed the retrieval method to the existing localization pipeline. The code for MegaLoc is available at https://github.com/gmberton/MegaLoc

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2502.17237")
get_code_for_paper("2502.17237")
have("2502.17237")

Connect an agent — have() is free.