Deep Convolutional Neural Networks as Generic Feature Extractors

Lars Hertel; Erhardt Barth; Thomas Kaster; Thomas Martinetz

doi:10.1109/IJCNN.2015.7280683

Deep Convolutional Neural Networks as Generic Feature Extractors

Lars Hertel, Erhardt Barth, Thomas Kaster, Thomas Martinetz

Abstract

Recognizing objects in natural images is an intricate problem involving multiple conflicting objectives. Deep convolutional neural networks, trained on large datasets, achieve convincing results and are currently the state-of-the-art approach for this task. However, the long time needed to train such deep networks is a major drawback. We tackled this problem by reusing a previously trained network. For this purpose, we first trained a deep convolutional network on the ILSVRC-12 dataset. We then maintained the learned convolution kernels and only retrained the classification part on different datasets. Using this approach, we achieved an accuracy of 67.68% on CIFAR-100, compared to the previous state-of-the-art result of 65.43%. Furthermore, our findings indicate that convolutional networks are able to learn generic feature extractors that can be used for different tasks.

Original language	English
Title of host publication	2015 International Joint Conference on Neural Networks (IJCNN)
Number of pages	4
Publisher	IEEE
Publication date	28.09.2015
ISBN (Print)	978-1-4799-1961-1
ISBN (Electronic)	978-1-4799-1960-4
DOIs	https://doi.org/10.1109/IJCNN.2015.7280683
Publication status	Published - 28.09.2015
Event	International Joint Conference on Neural Networks 2015 - Killarney, Ireland Duration: 12.07.2015 → 17.07.2015 https://ieeexplore.ieee.org/document/7280683/

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being
SDG 9 Industry, Innovation, and Infrastructure

Access to Document

10.1109/IJCNN.2015.7280683Licence: Unspecified

Cite this

@inproceedings{4b0ebe91e60244e4925cbdf125b2982c,

title = "Deep Convolutional Neural Networks as Generic Feature Extractors",

abstract = "Recognizing objects in natural images is an intricate problem involving multiple conflicting objectives. Deep convolutional neural networks, trained on large datasets, achieve convincing results and are currently the state-of-the-art approach for this task. However, the long time needed to train such deep networks is a major drawback. We tackled this problem by reusing a previously trained network. For this purpose, we first trained a deep convolutional network on the ILSVRC-12 dataset. We then maintained the learned convolution kernels and only retrained the classification part on different datasets. Using this approach, we achieved an accuracy of 67.68\% on CIFAR-100, compared to the previous state-of-the-art result of 65.43\%. Furthermore, our findings indicate that convolutional networks are able to learn generic feature extractors that can be used for different tasks.",

author = "Lars Hertel and Erhardt Barth and Thomas Kaster and Thomas Martinetz",

year = "2015",

month = sep,

day = "28",

doi = "10.1109/IJCNN.2015.7280683",

language = "English",

isbn = "978-1-4799-1961-1 ",

series = "IEEE Conference Proceedings",

publisher = "IEEE",

booktitle = "2015 International Joint Conference on Neural Networks (IJCNN)",

note = "International Joint Conference on Neural Networks 2015 ; Conference date: 12-07-2015 Through 17-07-2015",

url = "https://ieeexplore.ieee.org/document/7280683/",

}

TY - GEN

T1 - Deep Convolutional Neural Networks as Generic Feature Extractors

AU - Hertel, Lars

AU - Barth, Erhardt

AU - Kaster, Thomas

AU - Martinetz, Thomas

PY - 2015/9/28

Y1 - 2015/9/28

N2 - Recognizing objects in natural images is an intricate problem involving multiple conflicting objectives. Deep convolutional neural networks, trained on large datasets, achieve convincing results and are currently the state-of-the-art approach for this task. However, the long time needed to train such deep networks is a major drawback. We tackled this problem by reusing a previously trained network. For this purpose, we first trained a deep convolutional network on the ILSVRC-12 dataset. We then maintained the learned convolution kernels and only retrained the classification part on different datasets. Using this approach, we achieved an accuracy of 67.68% on CIFAR-100, compared to the previous state-of-the-art result of 65.43%. Furthermore, our findings indicate that convolutional networks are able to learn generic feature extractors that can be used for different tasks.

AB - Recognizing objects in natural images is an intricate problem involving multiple conflicting objectives. Deep convolutional neural networks, trained on large datasets, achieve convincing results and are currently the state-of-the-art approach for this task. However, the long time needed to train such deep networks is a major drawback. We tackled this problem by reusing a previously trained network. For this purpose, we first trained a deep convolutional network on the ILSVRC-12 dataset. We then maintained the learned convolution kernels and only retrained the classification part on different datasets. Using this approach, we achieved an accuracy of 67.68% on CIFAR-100, compared to the previous state-of-the-art result of 65.43%. Furthermore, our findings indicate that convolutional networks are able to learn generic feature extractors that can be used for different tasks.

U2 - 10.1109/IJCNN.2015.7280683

DO - 10.1109/IJCNN.2015.7280683

M3 - Conference contribution

SN - 978-1-4799-1961-1

T3 - IEEE Conference Proceedings

BT - 2015 International Joint Conference on Neural Networks (IJCNN)

PB - IEEE

T2 - International Joint Conference on Neural Networks 2015

Y2 - 12 July 2015 through 17 July 2015

ER -

Deep Convolutional Neural Networks as Generic Feature Extractors

Abstract

UN SDGs

Access to Document

Fingerprint

Cite this