Deep-Learning Approach to Topographical Image Analysis

Provider

Ministry of Education, Youth and Sports

Programme

INTER-EXCELLENCE

Call for proposals

INTER-EXCELLENCE 21 (SMSM2019LTAIZ)

Main participants

Vysoké učení technické v Brně / Fakulta informačních technologií

Contest type

VS - Public tender

Contract ID

MSMT-22782/2019-2

Project name in Czech

Topografická analýza obrazu s využitím metod hlubokého učení

Annotation in Czech

Cílem projektu je výzkum zcela nových přístupů na bázi hlubokého učení (angl. deep learning, DL) k fúzi multimodálních zdrojů dat, které jsou vhodné pro vizuální geo-lokalizaci. Jedná se zejména o fotografie/videa pořízené běžnou kamerou nebo mobilním zařízením, 3D digitální modely terénu, syntetické (renderované) obrazy, příp. informace o hloubce.

R&D category

ZV - Basic research

OECD FORD - main branch

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

OECD FORD - secondary branch

—

OECD FORD - another secondary branch

—

CEP - equivalent branches <br>(according to the <a href="http://www.vyzkum.cz/storage/att/E6EF7938F0E854BAE520AC119FB22E8D/Prevodnik_oboru_Frascati.pdf">converter</a>)

AF - Documentation, librarianship, work with information<br>BC - Theory and management systems<br>BD - Information theory<br>IN - Informatics

Provider evaluation

V - Vynikající výsledky projektu (s mezinárodním významem atd.)

Project results evaluation

"The project Deep-Learning Approach to Topographical Image Analysis focused on research of novel methods of visual camera localization based on multimodal data registration using current machine learning approaches. In the second half of the project (2021-22), we concentrated on methods for camera pose estimation in the natural environments. A new localization method Crosslocate was presented at the WACV22 conference, and an efficient method for horizon curve detection was introduced at the IJCNN21 conference. Research on the perception of 3D terrain models and other synthetically generated data also continued. Two new perceptual metrics for estimating the realism of 3D terrain, and tree models (ICTree) have been proposed and published in ACM TAP and at SIGGRAPH Asia 21, respectively. Two research datasets of terrain and botanical tree models with perceptual user scores obtained in large-scale user experiments were published as well.

Realization period - beginning

Jul 1, 2019

Realization period - end

Jun 30, 2022

Project status

U - Finished project

Latest support payment

Feb 24, 2022

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Data delivery code

CEP23-MSM-LT-U

Data delivery date

Jun 30, 2023

Total approved costs

4,478 thou. CZK

Public financial support

3,828 thou. CZK

Other public sources

0 thou. CZK

Non public and foreign sources

0 thou. CZK