Novel scene recognition using traindetector

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F62690094%3A18450%2F19%3A50015952" target="_blank" >RIV/62690094:18450/19:50015952 - isvavai.cz</a>

Výsledek na webu

<a href="http://dx.doi.org/10.1007/978-3-030-13469-3_59" target="_blank" >http://dx.doi.org/10.1007/978-3-030-13469-3_59</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/978-3-030-13469-3_59" target="_blank" >10.1007/978-3-030-13469-3_59</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Novel scene recognition using traindetector

Popis výsledku v původním jazyce

Our ability to process the image keeps improving day by day, since the introduction of deep learning. Lastly, this contributed to the advance of object recognition through a Convolutional neural network and Place recognition, which is our concern in this paper. Through this research, it was observed a complexity in the extraction of the correct and relevant features for scene recognition. To address this issue, we extracted at the pixel level several subareas which contain more color intensity than other parts, and we went through each image once to build the feature representation of it. We also noticed that several available models based on Convolution Neural Network requires a Graphics Processing Units (GPU) for their implementation and are difficult to train. We propose in this paper, a novel Scene Recognition method using Single-Shot-Detector (SSD), Multi-modal Local-Receptive-Field (MM-LRF) and Extreme-Learning-Machine (ELM) that we named TrainDetector. It outperforms the state-of-the-art techniques when we apply it to three well-known scene recognition Datasets.

Název v anglickém jazyce

Novel scene recognition using traindetector

Popis výsledku anglicky

Our ability to process the image keeps improving day by day, since the introduction of deep learning. Lastly, this contributed to the advance of object recognition through a Convolutional neural network and Place recognition, which is our concern in this paper. Through this research, it was observed a complexity in the extraction of the correct and relevant features for scene recognition. To address this issue, we extracted at the pixel level several subareas which contain more color intensity than other parts, and we went through each image once to build the feature representation of it. We also noticed that several available models based on Convolution Neural Network requires a Graphics Processing Units (GPU) for their implementation and are difficult to train. We propose in this paper, a novel Scene Recognition method using Single-Shot-Detector (SSD), Multi-modal Local-Receptive-Field (MM-LRF) and Extreme-Learning-Machine (ELM) that we named TrainDetector. It outperforms the state-of-the-art techniques when we apply it to three well-known scene recognition Datasets.

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

2019

Kód důvěrnosti údajů

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

Název statě ve sborníku

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ISBN

978-3-030-13468-6

ISSN

0302-9743

e-ISSN

—

Počet stran výsledku

9

Strana od-do

504-512

Název nakladatele

Springer Verlag

Místo vydání

Berlin

Místo konání akce

Madrid

Datum konání akce

19. 11. 2018

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—