Analytics of deep neural network in change detection

研究成果: 著書/レポートタイプへの貢献会議での発言

5 引用 (Scopus)

抄録

Recently, deep neural networks (DNNs) have demonstrated excellent performance for change detection. The DNN-based background subtraction automatically discovers background features from datasets and outperforms traditional background modeling based on handcraft features and/or subtraction strategies. Most researchers mainly discuss the accuracy of foreground detection and do not analyze how and why the DNN works well for change detection tasks. It is necessary to understand what the DNN learns as background features in order to discuss the potential of the DNN in background subtraction. In this paper, we focus on the filters in the first convolution layer and the activations of neurons in the last fully connected layer to understand the behavior of the DNN. From the experiment, we found that 1) the first layer performs the role of background subtraction using several filters, and 2) the last layer categorizes some background changes into a group without supervised signals. These findings suggest the possibility of a new background modeling strategy based on data-driven extracted features.

元の言語英語
ホスト出版物のタイトル2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017
出版者Institute of Electrical and Electronics Engineers Inc.
ISBN(電子版)9781538629390
DOI
出版物ステータス出版済み - 10 20 2017
イベント14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017 - Lecce, イタリア
継続期間: 8 29 20179 1 2017

出版物シリーズ

名前2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017

その他

その他14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017
イタリア
Lecce
期間8/29/179/1/17

Fingerprint

Convolution
Neurons
Deep neural networks
Chemical activation
Experiments

All Science Journal Classification (ASJC) codes

  • Computer Vision and Pattern Recognition
  • Signal Processing

これを引用

Minematsu, T., Shimada, A., & Taniguchi, R-I. (2017). Analytics of deep neural network in change detection. : 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017 [8078550] (2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/AVSS.2017.8078550

Analytics of deep neural network in change detection. / Minematsu, Tsubasa; Shimada, Atsushi; Taniguchi, Rin-Ichiro.

2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017. Institute of Electrical and Electronics Engineers Inc., 2017. 8078550 (2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017).

研究成果: 著書/レポートタイプへの貢献会議での発言

Minematsu, T, Shimada, A & Taniguchi, R-I 2017, Analytics of deep neural network in change detection. : 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017., 8078550, 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017, Institute of Electrical and Electronics Engineers Inc., 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017, Lecce, イタリア, 8/29/17. https://doi.org/10.1109/AVSS.2017.8078550
Minematsu T, Shimada A, Taniguchi R-I. Analytics of deep neural network in change detection. : 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017. Institute of Electrical and Electronics Engineers Inc. 2017. 8078550. (2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017). https://doi.org/10.1109/AVSS.2017.8078550
Minematsu, Tsubasa ; Shimada, Atsushi ; Taniguchi, Rin-Ichiro. / Analytics of deep neural network in change detection. 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017. Institute of Electrical and Electronics Engineers Inc., 2017. (2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017).
@inproceedings{18a4f3f372df4916b973b7bdcbab5b2f,
title = "Analytics of deep neural network in change detection",
abstract = "Recently, deep neural networks (DNNs) have demonstrated excellent performance for change detection. The DNN-based background subtraction automatically discovers background features from datasets and outperforms traditional background modeling based on handcraft features and/or subtraction strategies. Most researchers mainly discuss the accuracy of foreground detection and do not analyze how and why the DNN works well for change detection tasks. It is necessary to understand what the DNN learns as background features in order to discuss the potential of the DNN in background subtraction. In this paper, we focus on the filters in the first convolution layer and the activations of neurons in the last fully connected layer to understand the behavior of the DNN. From the experiment, we found that 1) the first layer performs the role of background subtraction using several filters, and 2) the last layer categorizes some background changes into a group without supervised signals. These findings suggest the possibility of a new background modeling strategy based on data-driven extracted features.",
author = "Tsubasa Minematsu and Atsushi Shimada and Rin-Ichiro Taniguchi",
year = "2017",
month = "10",
day = "20",
doi = "10.1109/AVSS.2017.8078550",
language = "English",
series = "2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
booktitle = "2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017",
address = "United States",

}

TY - GEN

T1 - Analytics of deep neural network in change detection

AU - Minematsu, Tsubasa

AU - Shimada, Atsushi

AU - Taniguchi, Rin-Ichiro

PY - 2017/10/20

Y1 - 2017/10/20

N2 - Recently, deep neural networks (DNNs) have demonstrated excellent performance for change detection. The DNN-based background subtraction automatically discovers background features from datasets and outperforms traditional background modeling based on handcraft features and/or subtraction strategies. Most researchers mainly discuss the accuracy of foreground detection and do not analyze how and why the DNN works well for change detection tasks. It is necessary to understand what the DNN learns as background features in order to discuss the potential of the DNN in background subtraction. In this paper, we focus on the filters in the first convolution layer and the activations of neurons in the last fully connected layer to understand the behavior of the DNN. From the experiment, we found that 1) the first layer performs the role of background subtraction using several filters, and 2) the last layer categorizes some background changes into a group without supervised signals. These findings suggest the possibility of a new background modeling strategy based on data-driven extracted features.

AB - Recently, deep neural networks (DNNs) have demonstrated excellent performance for change detection. The DNN-based background subtraction automatically discovers background features from datasets and outperforms traditional background modeling based on handcraft features and/or subtraction strategies. Most researchers mainly discuss the accuracy of foreground detection and do not analyze how and why the DNN works well for change detection tasks. It is necessary to understand what the DNN learns as background features in order to discuss the potential of the DNN in background subtraction. In this paper, we focus on the filters in the first convolution layer and the activations of neurons in the last fully connected layer to understand the behavior of the DNN. From the experiment, we found that 1) the first layer performs the role of background subtraction using several filters, and 2) the last layer categorizes some background changes into a group without supervised signals. These findings suggest the possibility of a new background modeling strategy based on data-driven extracted features.

UR - http://www.scopus.com/inward/record.url?scp=85039919660&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85039919660&partnerID=8YFLogxK

U2 - 10.1109/AVSS.2017.8078550

DO - 10.1109/AVSS.2017.8078550

M3 - Conference contribution

AN - SCOPUS:85039919660

T3 - 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017

BT - 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance, AVSS 2017

PB - Institute of Electrical and Electronics Engineers Inc.

ER -