2 Citations (Scopus)
268 Downloads (Pure)

Abstract

Object detection can be difficult due to challenges such as variations in objects both inter- and intra-class. Additionally, variations can also be present between images. Based on this, research was conducted into creating an ensemble of Region-based Fully Convolutional Networks (R-FCN) object detectors. Ensemble strategies explored were firstly data sampling and selection and secondly combination strategies. Data sampling and selection aimed to create different subsets of data with respect to object size and image quality such that expert R-FCN ensemble members could be trained. Two combination strategies were explored for combining the individual member detections into an ensemble result, namely average and a weighted average. R-FCNs were trained and tested on the PASCAL VOC benchmark object detection dataset. Results proved positive with an increase in Average Precision (AP), compared to state-of-the-art similar systems, when ensemble members were combined appropriately.
Original languageEnglish
Title of host publication International Joint Conference on Computational Intelligence
Volume1
PublisherSCITEPRESS Digital Library
Publication date2017
Pages110-120
ISBN (Print)978-989-758-274-5
DOIs
Publication statusPublished - 2017
EventInternational Joint Conference on Computational Intelligence - Funchal, Portugal
Duration: 1 Nov 20173 Nov 2017
Conference number: 9
http://www.ijcci.org/

Conference

ConferenceInternational Joint Conference on Computational Intelligence
Number9
CountryPortugal
CityFunchal
Period01/11/201703/11/2017
Internet address

Fingerprint

sampling
volatile organic compound
detection
detector
state of the art

Keywords

  • Convolutional Neural Networks
  • Ensemble Learning
  • Object Detection
  • Image Quality Assessment

Cite this

Rasmussen, C. B., Nasrollahi, K., & Moeslund, T. B. (2017). R-FCN Object Detection Ensemble based on Object Resolution and Image Quality. In International Joint Conference on Computational Intelligence (Vol. 1, pp. 110-120). SCITEPRESS Digital Library. https://doi.org/10.5220/0006511301100120
Rasmussen, Christoffer Bøgelund ; Nasrollahi, Kamal ; Moeslund, Thomas B. / R-FCN Object Detection Ensemble based on Object Resolution and Image Quality. International Joint Conference on Computational Intelligence. Vol. 1 SCITEPRESS Digital Library, 2017. pp. 110-120
@inproceedings{1d95304ddbad45038e78759b23dcec56,
title = "R-FCN Object Detection Ensemble based on Object Resolution and Image Quality",
abstract = "Object detection can be difficult due to challenges such as variations in objects both inter- and intra-class. Additionally, variations can also be present between images. Based on this, research was conducted into creating an ensemble of Region-based Fully Convolutional Networks (R-FCN) object detectors. Ensemble strategies explored were firstly data sampling and selection and secondly combination strategies. Data sampling and selection aimed to create different subsets of data with respect to object size and image quality such that expert R-FCN ensemble members could be trained. Two combination strategies were explored for combining the individual member detections into an ensemble result, namely average and a weighted average. R-FCNs were trained and tested on the PASCAL VOC benchmark object detection dataset. Results proved positive with an increase in Average Precision (AP), compared to state-of-the-art similar systems, when ensemble members were combined appropriately.",
keywords = "Convolutional Neural Networks, Ensemble Learning, Object Detection, Image Quality Assessment",
author = "Rasmussen, {Christoffer B{\o}gelund} and Kamal Nasrollahi and Moeslund, {Thomas B.}",
year = "2017",
doi = "10.5220/0006511301100120",
language = "English",
isbn = "978-989-758-274-5",
volume = "1",
pages = "110--120",
booktitle = "International Joint Conference on Computational Intelligence",
publisher = "SCITEPRESS Digital Library",

}

Rasmussen, CB, Nasrollahi, K & Moeslund, TB 2017, R-FCN Object Detection Ensemble based on Object Resolution and Image Quality. in International Joint Conference on Computational Intelligence. vol. 1, SCITEPRESS Digital Library, pp. 110-120, International Joint Conference on Computational Intelligence, Funchal, Portugal, 01/11/2017. https://doi.org/10.5220/0006511301100120

R-FCN Object Detection Ensemble based on Object Resolution and Image Quality. / Rasmussen, Christoffer Bøgelund; Nasrollahi, Kamal; Moeslund, Thomas B.

International Joint Conference on Computational Intelligence. Vol. 1 SCITEPRESS Digital Library, 2017. p. 110-120.

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

TY - GEN

T1 - R-FCN Object Detection Ensemble based on Object Resolution and Image Quality

AU - Rasmussen, Christoffer Bøgelund

AU - Nasrollahi, Kamal

AU - Moeslund, Thomas B.

PY - 2017

Y1 - 2017

N2 - Object detection can be difficult due to challenges such as variations in objects both inter- and intra-class. Additionally, variations can also be present between images. Based on this, research was conducted into creating an ensemble of Region-based Fully Convolutional Networks (R-FCN) object detectors. Ensemble strategies explored were firstly data sampling and selection and secondly combination strategies. Data sampling and selection aimed to create different subsets of data with respect to object size and image quality such that expert R-FCN ensemble members could be trained. Two combination strategies were explored for combining the individual member detections into an ensemble result, namely average and a weighted average. R-FCNs were trained and tested on the PASCAL VOC benchmark object detection dataset. Results proved positive with an increase in Average Precision (AP), compared to state-of-the-art similar systems, when ensemble members were combined appropriately.

AB - Object detection can be difficult due to challenges such as variations in objects both inter- and intra-class. Additionally, variations can also be present between images. Based on this, research was conducted into creating an ensemble of Region-based Fully Convolutional Networks (R-FCN) object detectors. Ensemble strategies explored were firstly data sampling and selection and secondly combination strategies. Data sampling and selection aimed to create different subsets of data with respect to object size and image quality such that expert R-FCN ensemble members could be trained. Two combination strategies were explored for combining the individual member detections into an ensemble result, namely average and a weighted average. R-FCNs were trained and tested on the PASCAL VOC benchmark object detection dataset. Results proved positive with an increase in Average Precision (AP), compared to state-of-the-art similar systems, when ensemble members were combined appropriately.

KW - Convolutional Neural Networks

KW - Ensemble Learning

KW - Object Detection

KW - Image Quality Assessment

U2 - 10.5220/0006511301100120

DO - 10.5220/0006511301100120

M3 - Article in proceeding

SN - 978-989-758-274-5

VL - 1

SP - 110

EP - 120

BT - International Joint Conference on Computational Intelligence

PB - SCITEPRESS Digital Library

ER -

Rasmussen CB, Nasrollahi K, Moeslund TB. R-FCN Object Detection Ensemble based on Object Resolution and Image Quality. In International Joint Conference on Computational Intelligence. Vol. 1. SCITEPRESS Digital Library. 2017. p. 110-120 https://doi.org/10.5220/0006511301100120