Learning domain specific features using convolutional autoencoder: A vein authentication case study using siamese triplet loss network

Manish Agnihotri, Aditya Rathod, Daksh Thapar, Gaurav Jaswal, Kamlesh Tiwari, Aditya Nigam

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

Abstract

Recently, deep hierarchically learned models (such as CNN) have achieved superior performance in various computer vision tasks but limited attention has been paid to biometrics till now. This is major because of the number of samples available in biometrics are limited and are not enough to train CNN efficiently. However, deep learning often requires a lot of training data because of the huge number of parameters to be tuned by the learning algorithm. How about designing an end-to-end deep learning network to match the biometric features when the number of training samples is limited? To address this problem, we propose a new way to design an end-to-end deep neural network that works in two major steps: first an auto-encoder has been trained for learning domain specific features followed by a Siamese network trained via. triplet loss function for matching. A publicly available vein image data set has been utilized as a case study to justify our proposal. We observed that transformations learned from such a network provide domain specific and most discriminative vascular features. Subsequently, the corresponding traits are matched using multimodal pipelined end-to-end network in which the convolutional layers are pre-trained in an unsupervised fashion as an autoencoder. Thorough experimental studies suggest that the proposed framework consistently outperforms several state-of-the-art vein recognition approaches.

Original languageEnglish
Title of host publicationICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods
EditorsAna Fred, Maria De Marsico, Gabriella Sanniti di Baja
PublisherSciTePress
Pages778-785
Number of pages8
ISBN (Electronic)9789897583513
Publication statusPublished - 01-01-2019
Externally publishedYes
Event8th International Conference on Pattern Recognition Applications and Methods, ICPRAM 2019 - Prague, Czech Republic
Duration: 19-02-201921-02-2019

Publication series

NameICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods

Conference

Conference8th International Conference on Pattern Recognition Applications and Methods, ICPRAM 2019
CountryCzech Republic
CityPrague
Period19-02-1921-02-19

Fingerprint

Biometrics
Authentication
Learning algorithms
Computer vision
Deep learning

All Science Journal Classification (ASJC) codes

  • Computer Vision and Pattern Recognition

Cite this

Agnihotri, M., Rathod, A., Thapar, D., Jaswal, G., Tiwari, K., & Nigam, A. (2019). Learning domain specific features using convolutional autoencoder: A vein authentication case study using siamese triplet loss network. In A. Fred, M. De Marsico, & G. S. di Baja (Eds.), ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods (pp. 778-785). (ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods). SciTePress.
Agnihotri, Manish ; Rathod, Aditya ; Thapar, Daksh ; Jaswal, Gaurav ; Tiwari, Kamlesh ; Nigam, Aditya. / Learning domain specific features using convolutional autoencoder : A vein authentication case study using siamese triplet loss network. ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods. editor / Ana Fred ; Maria De Marsico ; Gabriella Sanniti di Baja. SciTePress, 2019. pp. 778-785 (ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods).
@inproceedings{e90f15edc07f45cf9d60d6c07bee69d6,
title = "Learning domain specific features using convolutional autoencoder: A vein authentication case study using siamese triplet loss network",
abstract = "Recently, deep hierarchically learned models (such as CNN) have achieved superior performance in various computer vision tasks but limited attention has been paid to biometrics till now. This is major because of the number of samples available in biometrics are limited and are not enough to train CNN efficiently. However, deep learning often requires a lot of training data because of the huge number of parameters to be tuned by the learning algorithm. How about designing an end-to-end deep learning network to match the biometric features when the number of training samples is limited? To address this problem, we propose a new way to design an end-to-end deep neural network that works in two major steps: first an auto-encoder has been trained for learning domain specific features followed by a Siamese network trained via. triplet loss function for matching. A publicly available vein image data set has been utilized as a case study to justify our proposal. We observed that transformations learned from such a network provide domain specific and most discriminative vascular features. Subsequently, the corresponding traits are matched using multimodal pipelined end-to-end network in which the convolutional layers are pre-trained in an unsupervised fashion as an autoencoder. Thorough experimental studies suggest that the proposed framework consistently outperforms several state-of-the-art vein recognition approaches.",
author = "Manish Agnihotri and Aditya Rathod and Daksh Thapar and Gaurav Jaswal and Kamlesh Tiwari and Aditya Nigam",
year = "2019",
month = "1",
day = "1",
language = "English",
series = "ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods",
publisher = "SciTePress",
pages = "778--785",
editor = "Ana Fred and {De Marsico}, Maria and {di Baja}, {Gabriella Sanniti}",
booktitle = "ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods",

}

Agnihotri, M, Rathod, A, Thapar, D, Jaswal, G, Tiwari, K & Nigam, A 2019, Learning domain specific features using convolutional autoencoder: A vein authentication case study using siamese triplet loss network. in A Fred, M De Marsico & GS di Baja (eds), ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods. ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods, SciTePress, pp. 778-785, 8th International Conference on Pattern Recognition Applications and Methods, ICPRAM 2019, Prague, Czech Republic, 19-02-19.

Learning domain specific features using convolutional autoencoder : A vein authentication case study using siamese triplet loss network. / Agnihotri, Manish; Rathod, Aditya; Thapar, Daksh; Jaswal, Gaurav; Tiwari, Kamlesh; Nigam, Aditya.

ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods. ed. / Ana Fred; Maria De Marsico; Gabriella Sanniti di Baja. SciTePress, 2019. p. 778-785 (ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods).

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

TY - GEN

T1 - Learning domain specific features using convolutional autoencoder

T2 - A vein authentication case study using siamese triplet loss network

AU - Agnihotri, Manish

AU - Rathod, Aditya

AU - Thapar, Daksh

AU - Jaswal, Gaurav

AU - Tiwari, Kamlesh

AU - Nigam, Aditya

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Recently, deep hierarchically learned models (such as CNN) have achieved superior performance in various computer vision tasks but limited attention has been paid to biometrics till now. This is major because of the number of samples available in biometrics are limited and are not enough to train CNN efficiently. However, deep learning often requires a lot of training data because of the huge number of parameters to be tuned by the learning algorithm. How about designing an end-to-end deep learning network to match the biometric features when the number of training samples is limited? To address this problem, we propose a new way to design an end-to-end deep neural network that works in two major steps: first an auto-encoder has been trained for learning domain specific features followed by a Siamese network trained via. triplet loss function for matching. A publicly available vein image data set has been utilized as a case study to justify our proposal. We observed that transformations learned from such a network provide domain specific and most discriminative vascular features. Subsequently, the corresponding traits are matched using multimodal pipelined end-to-end network in which the convolutional layers are pre-trained in an unsupervised fashion as an autoencoder. Thorough experimental studies suggest that the proposed framework consistently outperforms several state-of-the-art vein recognition approaches.

AB - Recently, deep hierarchically learned models (such as CNN) have achieved superior performance in various computer vision tasks but limited attention has been paid to biometrics till now. This is major because of the number of samples available in biometrics are limited and are not enough to train CNN efficiently. However, deep learning often requires a lot of training data because of the huge number of parameters to be tuned by the learning algorithm. How about designing an end-to-end deep learning network to match the biometric features when the number of training samples is limited? To address this problem, we propose a new way to design an end-to-end deep neural network that works in two major steps: first an auto-encoder has been trained for learning domain specific features followed by a Siamese network trained via. triplet loss function for matching. A publicly available vein image data set has been utilized as a case study to justify our proposal. We observed that transformations learned from such a network provide domain specific and most discriminative vascular features. Subsequently, the corresponding traits are matched using multimodal pipelined end-to-end network in which the convolutional layers are pre-trained in an unsupervised fashion as an autoencoder. Thorough experimental studies suggest that the proposed framework consistently outperforms several state-of-the-art vein recognition approaches.

UR - http://www.scopus.com/inward/record.url?scp=85064615839&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85064615839&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:85064615839

T3 - ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods

SP - 778

EP - 785

BT - ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods

A2 - Fred, Ana

A2 - De Marsico, Maria

A2 - di Baja, Gabriella Sanniti

PB - SciTePress

ER -

Agnihotri M, Rathod A, Thapar D, Jaswal G, Tiwari K, Nigam A. Learning domain specific features using convolutional autoencoder: A vein authentication case study using siamese triplet loss network. In Fred A, De Marsico M, di Baja GS, editors, ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods. SciTePress. 2019. p. 778-785. (ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods).