A Novel Aerial Dataset for Scene Classification Annotated Using OSM for Learning Deep CNNs

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21230%2F18%3A00322143" target="_blank" >RIV/68407700:21230/18:00322143 - isvavai.cz</a>

Výsledek na webu

<a href="http://radio.feld.cvut.cz/conf/poster/proceedings/Poster_2018/Section_IC/IC_045_Kunc.pdf" target="_blank" >http://radio.feld.cvut.cz/conf/poster/proceedings/Poster_2018/Section_IC/IC_045_Kunc.pdf</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

A Novel Aerial Dataset for Scene Classification Annotated Using OSM for Learning Deep CNNs

Popis výsledku v původním jazyce

Remote sensing data are getting cheaper and thus tools allowing analysis of large quantities of the data are needed. One of the commonly used tools for automation of remote sensing data are neural networks. However, despite the rising availability of the data, there is no suitable dataset for learning neural networks. This paper introduces novel aerial image datasets that were automatically annotated using labels from OpenStreetMap. The largest of the datasets contains 52,596 400x400 px images divided into 44 classes. These datasets were used for learning a deep state-of-the-art neural network for image classification as the size of the datasets allows to learn such network from scratch which was difficult with currently available datasets. The classification performance of the neural networks represents the baseline performance for the presented datasets and was further analyzed using the gradCAM visualization method.

Název v anglickém jazyce

A Novel Aerial Dataset for Scene Classification Annotated Using OSM for Learning Deep CNNs

Popis výsledku anglicky

Remote sensing data are getting cheaper and thus tools allowing analysis of large quantities of the data are needed. One of the commonly used tools for automation of remote sensing data are neural networks. However, despite the rising availability of the data, there is no suitable dataset for learning neural networks. This paper introduces novel aerial image datasets that were automatically annotated using labels from OpenStreetMap. The largest of the datasets contains 52,596 400x400 px images divided into 44 classes. These datasets were used for learning a deep state-of-the-art neural network for image classification as the size of the datasets allows to learn such network from scratch which was difficult with currently available datasets. The classification performance of the neural networks represents the baseline performance for the presented datasets and was further analyzed using the gradCAM visualization method.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

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

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2018

Kód důvěrnosti údajů

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

Název statě ve sborníku

Proceedings of the International Student Scientific Conference Poster – 22/2018

ISBN

978-80-01-06428-3

ISSN

—

e-ISSN

—

Počet stran výsledku

5

Strana od-do

1-5

Název nakladatele

Czech Technical University in Prague

Místo vydání

Praha

Místo konání akce

Praha

Datum konání akce

10. 5. 2018

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—