You Are Here: Geolocation by Embedding Maps and Images

Obed N Samano Abonce; Mengjie Zhou; Andrew D Calway

You Are Here: Geolocation by Embedding Maps and Images

Obed N Samano Abonce, Mengjie Zhou, Andrew D Calway

Research output: Contribution to conference › Conference Paper › peer-review

196 Downloads (Pure)

Abstract

We present a novel approach to geolocalising panoramic images on a 2-D cartographic map based on learning a low dimensional embedded space, which allows a comparison between an image captured at a location and local neighbourhoods of the map. The representation is not sufficiently discriminatory to allow localisation from a single image, but when concatenated along a route, localisation converges quickly, with over 90% accuracy being achieved for routes of around 200m in length when using Google Street View and Open Street Map data. The method generalises a previous fixed semantic feature based approach and achieves significantly higher localisation accuracy and faster convergence.

Original language	English
Number of pages	17
Publication status	Published - 28 Aug 2020
Event	16th European Conference on Computer Vision - Online Duration: 23 Aug 2020 → 28 Aug 2020 https://eccv2020.eu

Conference

Conference	16th European Conference on Computer Vision
Abbreviated title	ECCV20
Period	23/08/20 → 28/08/20
Internet address	https://eccv2020.eu

Keywords

Geolocalisation
image-map embeddings
cross domain localisation
representation learning

Access to Document

Full-text PDF (author’s accepted manuscript)Accepted author manuscript, 4.65 MB

Handle.net

Persistent link

Professor Andrew Calway
- Andrew.Calwaybristol.acuk
- School of Computer Science - Professor of Computer Vision
- Visual Information Laboratory
Person: Academic , Member

Cite this

@conference{d80154fb123b4e9d9b4bdc751871e1a3,

title = "You Are Here: Geolocation by Embedding Maps and Images",

abstract = "We present a novel approach to geolocalising panoramic images on a 2-D cartographic map based on learning a low dimensional embedded space, which allows a comparison between an image captured at a location and local neighbourhoods of the map. The representation is not sufficiently discriminatory to allow localisation from a single image, but when concatenated along a route, localisation converges quickly, with over 90% accuracy being achieved for routes of around 200m in length when using Google Street View and Open Street Map data. The method generalises a previous fixed semantic feature based approach and achieves significantly higher localisation accuracy and faster convergence.",

keywords = "Geolocalisation, image-map embeddings, cross domain localisation, representation learning",

author = "{Samano Abonce}, {Obed N} and Mengjie Zhou and Calway, {Andrew D}",

year = "2020",

month = aug,

day = "28",

language = "English",

note = "16th European Conference on Computer Vision, ECCV20 ; Conference date: 23-08-2020 Through 28-08-2020",

url = "https://eccv2020.eu",

}

TY - CONF

T1 - You Are Here

T2 - 16th European Conference on Computer Vision

AU - Samano Abonce, Obed N

AU - Zhou, Mengjie

AU - Calway, Andrew D

PY - 2020/8/28

Y1 - 2020/8/28

N2 - We present a novel approach to geolocalising panoramic images on a 2-D cartographic map based on learning a low dimensional embedded space, which allows a comparison between an image captured at a location and local neighbourhoods of the map. The representation is not sufficiently discriminatory to allow localisation from a single image, but when concatenated along a route, localisation converges quickly, with over 90% accuracy being achieved for routes of around 200m in length when using Google Street View and Open Street Map data. The method generalises a previous fixed semantic feature based approach and achieves significantly higher localisation accuracy and faster convergence.

AB - We present a novel approach to geolocalising panoramic images on a 2-D cartographic map based on learning a low dimensional embedded space, which allows a comparison between an image captured at a location and local neighbourhoods of the map. The representation is not sufficiently discriminatory to allow localisation from a single image, but when concatenated along a route, localisation converges quickly, with over 90% accuracy being achieved for routes of around 200m in length when using Google Street View and Open Street Map data. The method generalises a previous fixed semantic feature based approach and achieves significantly higher localisation accuracy and faster convergence.

KW - Geolocalisation

KW - image-map embeddings

KW - cross domain localisation

KW - representation learning

M3 - Conference Paper

Y2 - 23 August 2020 through 28 August 2020

ER -

You Are Here: Geolocation by Embedding Maps and Images

Abstract

Conference

Keywords

Access to Document

Handle.net

Fingerprint

Profiles

Professor Andrew Calway

Cite this