TY - JOUR
T1 - Face identification using multi-layer perceptron and convolutional neural network
AU - Fithriasari, Kartika
AU - Nuraini, Ulfa Siti
N1 - Publisher Copyright:
© 2021 ICIC International
PY - 2021/2
Y1 - 2021/2
N2 - Algorithm that is often used for image processing is machine learning. One of machine learning is Neural Network, which has additional layer called Multi-Layer Perceptron (MLP). Besides, other technique is Convolutional Neural Network (CNN). In this paper, we compare MLP and CNN methods carried out in image processing for face identification case study. We used preprocessing to change RGB to grayscale and normalize it. We proposed to use image augmentation to get more data without taking some images again. Lack of images can make overfitting so image augmentation may reduce it. For training, we used Adam optimizer. It is more efficient to train images that have complicated patterns. The conclusion is that CNN method has better results, due to its higher accuracy, precision, sensitivity, and Fscore than MLP method.
AB - Algorithm that is often used for image processing is machine learning. One of machine learning is Neural Network, which has additional layer called Multi-Layer Perceptron (MLP). Besides, other technique is Convolutional Neural Network (CNN). In this paper, we compare MLP and CNN methods carried out in image processing for face identification case study. We used preprocessing to change RGB to grayscale and normalize it. We proposed to use image augmentation to get more data without taking some images again. Lack of images can make overfitting so image augmentation may reduce it. For training, we used Adam optimizer. It is more efficient to train images that have complicated patterns. The conclusion is that CNN method has better results, due to its higher accuracy, precision, sensitivity, and Fscore than MLP method.
KW - Convolutional neural network
KW - Face
KW - Image
KW - Multi-layer perceptron
UR - http://www.scopus.com/inward/record.url?scp=85099338193&partnerID=8YFLogxK
U2 - 10.24507/icicel.15.02.157
DO - 10.24507/icicel.15.02.157
M3 - Article
AN - SCOPUS:85099338193
SN - 1881-803X
VL - 15
SP - 157
EP - 164
JO - ICIC Express Letters
JF - ICIC Express Letters
IS - 2
ER -