Convolutional Neural Network as an Extractor Feature For Image Search

Sandi Rodiyansyah; Ardi Mardiana

Selected Papers from the 1st International Conference on Islam, Science and Technology, ICONISTECH-1 2019, 11-12 July 2019, Bandung, Indonesia

Research Article

Convolutional Neural Network as an Extractor Feature For Image Search

Download330 downloads

Cite: BibTeX Plain Text

@INPROCEEDINGS{10.4108/eai.11-7-2019.2297409,
    author={Sandi Fajar Rodiyansyah and Ardi  Mardiana},
    title={Convolutional Neural Network as an Extractor Feature For Image Search},
    proceedings={Selected Papers from the 1st International Conference on Islam, Science and Technology, ICONISTECH-1 2019, 11-12 July 2019, Bandung, Indonesia},
    publisher={EAI},
    proceedings_a={ICONISTECH-1},
    year={2020},
    month={11},
    keywords={deep learning; convolutional neural network; extractor feature; image search},
    doi={10.4108/eai.11-7-2019.2297409}
}

Sandi Fajar Rodiyansyah
Ardi Mardiana
Year: 2020
Convolutional Neural Network as an Extractor Feature For Image Search
ICONISTECH-1
EAI
DOI: 10.4108/eai.11-7-2019.2297409

Sandi Fajar Rodiyansyah¹^,*, Ardi Mardiana¹

1: Universitas Majalengka, Majalengka, West Java, Indonesia

*Contact email: rodiyansyah@unma.ac.id

Abstract

The deep learning method that can be used for image search was a convolutional neural network but there were many parameters and design decisions that were difficult to determine. In this study, a convolutional neural network method was used as an extractor feature for image search. Optimal feature extraction was performed at the last second fully connected layer (FC2) and used cosine distance as a distance metric with a threshold of 0.4. Producing the accuracy of model classification on the test data was 87.72%.

Keywords: deep learning; convolutional neural network; extractor feature; image search

Published: 2020-11-11
Publisher: EAI

: http://dx.doi.org/10.4108/eai.11-7-2019.2297409