Simultaneous object segmentation and recognition by merging CNN outputs from uniformly distributed multiple viewpoints

Yoshikatsu Nakajima, Hideo Saito

Research output: Contribution to journalArticle

Abstract

We propose a novel object recognition system that is able to (i) work in real-time while reconstructing segmented 3D maps and simultaneously recognize objects in a scene, (ii) manage various kinds of objects, including those with smooth surfaces and those with a large number of categories, utilizing a CNN for feature extraction, and (iii) maintain high accuracy no matter how the camera moves by distributing the viewpoints for each object uniformly and aggregating recognition results from each distributed viewpoint as the same weight. Through experiments, the advantages of our system with respect to current state-of-the-art object recognition approaches are demonstrated on the UW RGB-D Dataset and Scenes and on our own scenes prepared to verify the effectiveness of the Viewpoint-Class-based approach.

Original languageEnglish
Pages (from-to)1308-1316
Number of pages9
JournalIEICE Transactions on Information and Systems
VolumeE101D
Issue number5
DOIs
Publication statusPublished - 2018 May 1

Fingerprint

Object recognition
Merging
Feature extraction
Cameras
Experiments

Keywords

  • Convolutional neural network
  • Object recognition
  • Segmentation
  • SLAM

ASJC Scopus subject areas

  • Software
  • Hardware and Architecture
  • Computer Vision and Pattern Recognition
  • Electrical and Electronic Engineering
  • Artificial Intelligence

Cite this

Simultaneous object segmentation and recognition by merging CNN outputs from uniformly distributed multiple viewpoints. / Nakajima, Yoshikatsu; Saito, Hideo.

In: IEICE Transactions on Information and Systems, Vol. E101D, No. 5, 01.05.2018, p. 1308-1316.

Research output: Contribution to journalArticle

@article{4f00a3ce40d34758897946cdcefff07c,
title = "Simultaneous object segmentation and recognition by merging CNN outputs from uniformly distributed multiple viewpoints",
abstract = "We propose a novel object recognition system that is able to (i) work in real-time while reconstructing segmented 3D maps and simultaneously recognize objects in a scene, (ii) manage various kinds of objects, including those with smooth surfaces and those with a large number of categories, utilizing a CNN for feature extraction, and (iii) maintain high accuracy no matter how the camera moves by distributing the viewpoints for each object uniformly and aggregating recognition results from each distributed viewpoint as the same weight. Through experiments, the advantages of our system with respect to current state-of-the-art object recognition approaches are demonstrated on the UW RGB-D Dataset and Scenes and on our own scenes prepared to verify the effectiveness of the Viewpoint-Class-based approach.",
keywords = "Convolutional neural network, Object recognition, Segmentation, SLAM",
author = "Yoshikatsu Nakajima and Hideo Saito",
year = "2018",
month = "5",
day = "1",
doi = "10.1587/transinf.2017MVP0024",
language = "English",
volume = "E101D",
pages = "1308--1316",
journal = "IEICE Transactions on Information and Systems",
issn = "0916-8532",
publisher = "Maruzen Co., Ltd/Maruzen Kabushikikaisha",
number = "5",

}

TY - JOUR

T1 - Simultaneous object segmentation and recognition by merging CNN outputs from uniformly distributed multiple viewpoints

AU - Nakajima, Yoshikatsu

AU - Saito, Hideo

PY - 2018/5/1

Y1 - 2018/5/1

N2 - We propose a novel object recognition system that is able to (i) work in real-time while reconstructing segmented 3D maps and simultaneously recognize objects in a scene, (ii) manage various kinds of objects, including those with smooth surfaces and those with a large number of categories, utilizing a CNN for feature extraction, and (iii) maintain high accuracy no matter how the camera moves by distributing the viewpoints for each object uniformly and aggregating recognition results from each distributed viewpoint as the same weight. Through experiments, the advantages of our system with respect to current state-of-the-art object recognition approaches are demonstrated on the UW RGB-D Dataset and Scenes and on our own scenes prepared to verify the effectiveness of the Viewpoint-Class-based approach.

AB - We propose a novel object recognition system that is able to (i) work in real-time while reconstructing segmented 3D maps and simultaneously recognize objects in a scene, (ii) manage various kinds of objects, including those with smooth surfaces and those with a large number of categories, utilizing a CNN for feature extraction, and (iii) maintain high accuracy no matter how the camera moves by distributing the viewpoints for each object uniformly and aggregating recognition results from each distributed viewpoint as the same weight. Through experiments, the advantages of our system with respect to current state-of-the-art object recognition approaches are demonstrated on the UW RGB-D Dataset and Scenes and on our own scenes prepared to verify the effectiveness of the Viewpoint-Class-based approach.

KW - Convolutional neural network

KW - Object recognition

KW - Segmentation

KW - SLAM

UR - http://www.scopus.com/inward/record.url?scp=85046267796&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85046267796&partnerID=8YFLogxK

U2 - 10.1587/transinf.2017MVP0024

DO - 10.1587/transinf.2017MVP0024

M3 - Article

AN - SCOPUS:85046267796

VL - E101D

SP - 1308

EP - 1316

JO - IEICE Transactions on Information and Systems

JF - IEICE Transactions on Information and Systems

SN - 0916-8532

IS - 5

ER -