Unknown Landscape Identification with CNN Transfer Learning

Edoardo Serra; Ashish Sharma; Mikel Joaristi; Oxana Korzh

doi:10.1109/ASONAM.2018.8508357

Unknown Landscape Identification with CNN Transfer Learning

Edoardo Serra, Ashish Sharma, Mikel Joaristi, Oxana Korzh

Computer Science Department

Boise State University

Research output: Chapter in Book/Report/Conference proceeding › Chapter

14 Scopus citations

Abstract

Unknown landscape identification is the problem of identifying an unknown landscape from a set of already provided landscape images that are considered to be known. The aim of this work is to extract the intrinsic semantic of landscape images in order to automatically generalize concepts like a stadium, roads, a parking lot etc., and use this concept to identify unknown landscapes. This problem can be easily extended to many security applications. We propose two effective semi-supervised novelty detection approaches for the unknown landscape identification problem using Convolutional Neural Network (CNN) Transfer Learning. This is based on the use of pre-trained CNNs (i.e. already trained on large datasets) already containing general image knowledge that we transfer to our domain. Our best values of AUROC and Average Precision scores for the identification problem are 0.96 and 0.94, respectively. In addition, we statistically prove that our semi-supervised methods outperform the baseline.

Original language	American English
Title of host publication	ASONAM 2018: Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining: FAB 2018, FOSINT-SI 2018, HI-BI-BI 2018
Editors	Andrea Tagarelli, Chandan Reddy, Ulrik Brandes
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	813-820
Number of pages	8
ISBN (Electronic)	9781538660515
DOIs	https://doi.org/10.1109/ASONAM.2018.8508357
State	Published - 1 Jan 2018
Event	10th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018 - Barcelona, Spain Duration: 28 Aug 2018 → 31 Aug 2018

Publication series

Name	Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018

Conference

Conference	10th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018
Country/Territory	Spain
City	Barcelona
Period	28/08/18 → 31/08/18

Keywords

Gaussian mixture model
feature extraction
standards
support vector machines
task analysis
training

EGS Disciplines

Computer Sciences

Access to Document

10.1109/ASONAM.2018.8508357License: Unspecified

https://doi.org/10.1109/ASONAM.2018.8508357

Cite this

Serra, E., Sharma, A., Joaristi, M., & Korzh, O. (2018). Unknown Landscape Identification with CNN Transfer Learning. In A. Tagarelli, C. Reddy, & U. Brandes (Eds.), ASONAM 2018: Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining: FAB 2018, FOSINT-SI 2018, HI-BI-BI 2018 (pp. 813-820). (Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ASONAM.2018.8508357

Serra, Edoardo ; Sharma, Ashish ; Joaristi, Mikel et al. / Unknown Landscape Identification with CNN Transfer Learning. ASONAM 2018: Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining: FAB 2018, FOSINT-SI 2018, HI-BI-BI 2018. editor / Andrea Tagarelli ; Chandan Reddy ; Ulrik Brandes. Institute of Electrical and Electronics Engineers Inc., 2018. pp. 813-820 (Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018).

@inbook{b1709636eab745ffa1702da80fa527d0,

title = "Unknown Landscape Identification with CNN Transfer Learning",

abstract = " Unknown landscape identification is the problem of identifying an unknown landscape from a set of already provided landscape images that are considered to be known. The aim of this work is to extract the intrinsic semantic of landscape images in order to automatically generalize concepts like a stadium, roads, a parking lot etc., and use this concept to identify unknown landscapes. This problem can be easily extended to many security applications. We propose two effective semi-supervised novelty detection approaches for the unknown landscape identification problem using Convolutional Neural Network (CNN) Transfer Learning. This is based on the use of pre-trained CNNs (i.e. already trained on large datasets) already containing general image knowledge that we transfer to our domain. Our best values of AUROC and Average Precision scores for the identification problem are 0.96 and 0.94, respectively. In addition, we statistically prove that our semi-supervised methods outperform the baseline.",

keywords = "Gaussian mixture model, feature extraction, standards, support vector machines, task analysis, training",

author = "Edoardo Serra and Ashish Sharma and Mikel Joaristi and Oxana Korzh",

note = "Publisher Copyright: {\textcopyright} 2018 IEEE.; 10th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018 ; Conference date: 28-08-2018 Through 31-08-2018",

year = "2018",

month = jan,

day = "1",

doi = "10.1109/ASONAM.2018.8508357",

language = "American English",

series = "Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "813--820",

editor = "Andrea Tagarelli and Chandan Reddy and Ulrik Brandes",

booktitle = "ASONAM 2018: Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining: FAB 2018, FOSINT-SI 2018, HI-BI-BI 2018",

}

Serra, E, Sharma, A, Joaristi, M & Korzh, O 2018, Unknown Landscape Identification with CNN Transfer Learning. in A Tagarelli, C Reddy & U Brandes (eds), ASONAM 2018: Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining: FAB 2018, FOSINT-SI 2018, HI-BI-BI 2018. Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018, Institute of Electrical and Electronics Engineers Inc., pp. 813-820, 10th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018, Barcelona, Spain, 28/08/18. https://doi.org/10.1109/ASONAM.2018.8508357

Unknown Landscape Identification with CNN Transfer Learning. / Serra, Edoardo; Sharma, Ashish; Joaristi, Mikel et al.
ASONAM 2018: Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining: FAB 2018, FOSINT-SI 2018, HI-BI-BI 2018. ed. / Andrea Tagarelli; Chandan Reddy; Ulrik Brandes. Institute of Electrical and Electronics Engineers Inc., 2018. p. 813-820 (Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018).

Research output: Chapter in Book/Report/Conference proceeding › Chapter

TY - CHAP

T1 - Unknown Landscape Identification with CNN Transfer Learning

AU - Serra, Edoardo

AU - Sharma, Ashish

AU - Joaristi, Mikel

AU - Korzh, Oxana

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Unknown landscape identification is the problem of identifying an unknown landscape from a set of already provided landscape images that are considered to be known. The aim of this work is to extract the intrinsic semantic of landscape images in order to automatically generalize concepts like a stadium, roads, a parking lot etc., and use this concept to identify unknown landscapes. This problem can be easily extended to many security applications. We propose two effective semi-supervised novelty detection approaches for the unknown landscape identification problem using Convolutional Neural Network (CNN) Transfer Learning. This is based on the use of pre-trained CNNs (i.e. already trained on large datasets) already containing general image knowledge that we transfer to our domain. Our best values of AUROC and Average Precision scores for the identification problem are 0.96 and 0.94, respectively. In addition, we statistically prove that our semi-supervised methods outperform the baseline.

AB - Unknown landscape identification is the problem of identifying an unknown landscape from a set of already provided landscape images that are considered to be known. The aim of this work is to extract the intrinsic semantic of landscape images in order to automatically generalize concepts like a stadium, roads, a parking lot etc., and use this concept to identify unknown landscapes. This problem can be easily extended to many security applications. We propose two effective semi-supervised novelty detection approaches for the unknown landscape identification problem using Convolutional Neural Network (CNN) Transfer Learning. This is based on the use of pre-trained CNNs (i.e. already trained on large datasets) already containing general image knowledge that we transfer to our domain. Our best values of AUROC and Average Precision scores for the identification problem are 0.96 and 0.94, respectively. In addition, we statistically prove that our semi-supervised methods outperform the baseline.

KW - Gaussian mixture model

KW - feature extraction

KW - standards

KW - support vector machines

KW - task analysis

KW - training

UR - https://scholarworks.boisestate.edu/cs_facpubs/180

UR - https://doi.org/10.1109/ASONAM.2018.8508357

UR - http://www.scopus.com/inward/record.url?scp=85057310430&partnerID=8YFLogxK

U2 - 10.1109/ASONAM.2018.8508357

DO - 10.1109/ASONAM.2018.8508357

M3 - Chapter

T3 - Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018

SP - 813

EP - 820

BT - ASONAM 2018: Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining: FAB 2018, FOSINT-SI 2018, HI-BI-BI 2018

A2 - Tagarelli, Andrea

A2 - Reddy, Chandan

A2 - Brandes, Ulrik

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 10th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018

Y2 - 28 August 2018 through 31 August 2018

ER -

Serra E, Sharma A, Joaristi M, Korzh O. Unknown Landscape Identification with CNN Transfer Learning. In Tagarelli A, Reddy C, Brandes U, editors, ASONAM 2018: Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining: FAB 2018, FOSINT-SI 2018, HI-BI-BI 2018. Institute of Electrical and Electronics Engineers Inc. 2018. p. 813-820. (Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, ASONAM 2018). doi: 10.1109/ASONAM.2018.8508357

Unknown Landscape Identification with CNN Transfer Learning

Abstract

Publication series

Conference

Keywords

EGS Disciplines

Access to Document

Other files and links

Fingerprint

Cite this