Indoor Semantic Segmentation for Robot Navigating on Mobile

Wonsuk Kim, Junhee Seok

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

In recent years, there have been many successes of using Deep Convolutional Neural Networks (DCNNs) in the task of pixel-level classification (also called 'semantic image segmentation'). The advances in DCNN have led to the development of autonomous vehicles that can drive with no driver controls by using sensors like camera, LiDAR, etc. In this paper, we propose a practical method to implement autonomous indoor navigation based on semantic image segmentation using state-of-the-art performance model on mobile devices, especially Android devices. We apply a system called 'Mobile DeepLabv3', which uses atrous convolution when applying semantic image segmentation by using MobileNetV2 as a network backbone. The ADE20K dataset is used to train our models specific to indoor environments. Since this model is for robot navigating, we re-label 150 classes into 20 classes in order to easily classify obstacles and road. We evaluate the trade-offs between accuracy and computational complexity, as well as actual latency and the number of parameters of the trained models.

Original languageEnglish
Title of host publicationICUFN 2018 - 10th International Conference on Ubiquitous and Future Networks
PublisherIEEE Computer Society
Pages22-25
Number of pages4
Volume2018-July
ISBN (Print)9781538646465
DOIs
Publication statusPublished - 2018 Aug 14
Event10th International Conference on Ubiquitous and Future Networks, ICUFN 2018 - Prague, Czech Republic
Duration: 2018 Jul 32018 Jul 6

Other

Other10th International Conference on Ubiquitous and Future Networks, ICUFN 2018
CountryCzech Republic
CityPrague
Period18/7/318/7/6

Fingerprint

Semantics
Image segmentation
Robots
Neural networks
Convolution
Mobile devices
Labels
Computational complexity
Navigation
Pixels
Cameras
Sensors

Keywords

  • Atrous Convolution
  • Convolutional Neural Networks
  • Indoor Navigation
  • Semantic Segmentation

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Computer Science Applications
  • Hardware and Architecture

Cite this

Kim, W., & Seok, J. (2018). Indoor Semantic Segmentation for Robot Navigating on Mobile. In ICUFN 2018 - 10th International Conference on Ubiquitous and Future Networks (Vol. 2018-July, pp. 22-25). [8436956] IEEE Computer Society. https://doi.org/10.1109/ICUFN.2018.8436956

Indoor Semantic Segmentation for Robot Navigating on Mobile. / Kim, Wonsuk; Seok, Junhee.

ICUFN 2018 - 10th International Conference on Ubiquitous and Future Networks. Vol. 2018-July IEEE Computer Society, 2018. p. 22-25 8436956.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Kim, W & Seok, J 2018, Indoor Semantic Segmentation for Robot Navigating on Mobile. in ICUFN 2018 - 10th International Conference on Ubiquitous and Future Networks. vol. 2018-July, 8436956, IEEE Computer Society, pp. 22-25, 10th International Conference on Ubiquitous and Future Networks, ICUFN 2018, Prague, Czech Republic, 18/7/3. https://doi.org/10.1109/ICUFN.2018.8436956
Kim W, Seok J. Indoor Semantic Segmentation for Robot Navigating on Mobile. In ICUFN 2018 - 10th International Conference on Ubiquitous and Future Networks. Vol. 2018-July. IEEE Computer Society. 2018. p. 22-25. 8436956 https://doi.org/10.1109/ICUFN.2018.8436956
Kim, Wonsuk ; Seok, Junhee. / Indoor Semantic Segmentation for Robot Navigating on Mobile. ICUFN 2018 - 10th International Conference on Ubiquitous and Future Networks. Vol. 2018-July IEEE Computer Society, 2018. pp. 22-25
@inproceedings{64c383fe84ec47e786bcb586ca01955d,
title = "Indoor Semantic Segmentation for Robot Navigating on Mobile",
abstract = "In recent years, there have been many successes of using Deep Convolutional Neural Networks (DCNNs) in the task of pixel-level classification (also called 'semantic image segmentation'). The advances in DCNN have led to the development of autonomous vehicles that can drive with no driver controls by using sensors like camera, LiDAR, etc. In this paper, we propose a practical method to implement autonomous indoor navigation based on semantic image segmentation using state-of-the-art performance model on mobile devices, especially Android devices. We apply a system called 'Mobile DeepLabv3', which uses atrous convolution when applying semantic image segmentation by using MobileNetV2 as a network backbone. The ADE20K dataset is used to train our models specific to indoor environments. Since this model is for robot navigating, we re-label 150 classes into 20 classes in order to easily classify obstacles and road. We evaluate the trade-offs between accuracy and computational complexity, as well as actual latency and the number of parameters of the trained models.",
keywords = "Atrous Convolution, Convolutional Neural Networks, Indoor Navigation, Semantic Segmentation",
author = "Wonsuk Kim and Junhee Seok",
year = "2018",
month = "8",
day = "14",
doi = "10.1109/ICUFN.2018.8436956",
language = "English",
isbn = "9781538646465",
volume = "2018-July",
pages = "22--25",
booktitle = "ICUFN 2018 - 10th International Conference on Ubiquitous and Future Networks",
publisher = "IEEE Computer Society",

}

TY - GEN

T1 - Indoor Semantic Segmentation for Robot Navigating on Mobile

AU - Kim, Wonsuk

AU - Seok, Junhee

PY - 2018/8/14

Y1 - 2018/8/14

N2 - In recent years, there have been many successes of using Deep Convolutional Neural Networks (DCNNs) in the task of pixel-level classification (also called 'semantic image segmentation'). The advances in DCNN have led to the development of autonomous vehicles that can drive with no driver controls by using sensors like camera, LiDAR, etc. In this paper, we propose a practical method to implement autonomous indoor navigation based on semantic image segmentation using state-of-the-art performance model on mobile devices, especially Android devices. We apply a system called 'Mobile DeepLabv3', which uses atrous convolution when applying semantic image segmentation by using MobileNetV2 as a network backbone. The ADE20K dataset is used to train our models specific to indoor environments. Since this model is for robot navigating, we re-label 150 classes into 20 classes in order to easily classify obstacles and road. We evaluate the trade-offs between accuracy and computational complexity, as well as actual latency and the number of parameters of the trained models.

AB - In recent years, there have been many successes of using Deep Convolutional Neural Networks (DCNNs) in the task of pixel-level classification (also called 'semantic image segmentation'). The advances in DCNN have led to the development of autonomous vehicles that can drive with no driver controls by using sensors like camera, LiDAR, etc. In this paper, we propose a practical method to implement autonomous indoor navigation based on semantic image segmentation using state-of-the-art performance model on mobile devices, especially Android devices. We apply a system called 'Mobile DeepLabv3', which uses atrous convolution when applying semantic image segmentation by using MobileNetV2 as a network backbone. The ADE20K dataset is used to train our models specific to indoor environments. Since this model is for robot navigating, we re-label 150 classes into 20 classes in order to easily classify obstacles and road. We evaluate the trade-offs between accuracy and computational complexity, as well as actual latency and the number of parameters of the trained models.

KW - Atrous Convolution

KW - Convolutional Neural Networks

KW - Indoor Navigation

KW - Semantic Segmentation

UR - http://www.scopus.com/inward/record.url?scp=85052521968&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85052521968&partnerID=8YFLogxK

U2 - 10.1109/ICUFN.2018.8436956

DO - 10.1109/ICUFN.2018.8436956

M3 - Conference contribution

AN - SCOPUS:85052521968

SN - 9781538646465

VL - 2018-July

SP - 22

EP - 25

BT - ICUFN 2018 - 10th International Conference on Ubiquitous and Future Networks

PB - IEEE Computer Society

ER -