![](bg1.png)
Convolutional Neural Network Face Recognition for Lecturer
Attendance
Muhammad Rafi Muttaqin
1
, Anshorulloh Nur Aziz
1
, Dede Irmayanti
1
and Sumanto
2
1
Informatics Engineering Study Program, Wastukancana College of Technology, Purwakarta, Indonesia
2
Universitas Bina Sarana Informatika, Jakarta, Indonesia
fi
Keywords:
Face Recognition, MobileNet V2, Attendance.
Abstract:
Face recognition is a field of research that is widely used to solve various problems, but to apply face recogni-
tion requires high accuracy so that there are no errors in the system that applies face recognition. The purpose
of this research is how to use one of the architectures of the Convolutional Neural Network (CNN), namely
MobileNet v2 to perform the task of face recognition of STT Wastukancana lecturers. The data used is taken
from the social media of each lecturer, data sharing is done with the K-Fold Cross Validation method. Mo-
bileNet v2 architecture will perform classification tasks using different hyperparameter values to find the best
performance. From various patterns, the best accuracy is 85dropout of 0.3 to reduce overfitting. Data sharing
using K-Fold Cross Validation provides results that improve accuracy. The addition of a dropout layer reduces
overfitting of the model.
1 INTRODUCTION
A face is one way to recognize a person’s identity.
Humans can recognize someone’s name from look-
ing at their face, if they have known that person be-
fore. Many computer applications or systems that are
made require a person’s identity, and there are also
many ways to recognize that identity. Attendance sys-
tem is one of the examples. There are various ways
used in an attendance system, one of the simplest is
by signing on paper which is now used in the atten-
dance system for lecturers at STT Wastukancana. To
facilitate the attendance system, face recognition can
be applied to replace the manual signature process on
paper. Basically, face recognition is an image classi-
fication that is specialized for face classification only.
Convolutional neural network (CNN) is the most suit-
able model used for image classification, because it
has been specialized to separate and detect patterns in
input images, thus making this approach useful in the
field of face recognition(Farayola and Dureja, 2020).
There are various CNN architectures such as AlexNet,
GoogleNet, LeNet 5, or MobileNet. In this journal,
the author will use the MobileNet v2 architecture, be-
cause this model was developed for efficiency and
without sacrificing many resources (S. K. A. B. Singh,
2019). MobileNet is built using a deeply decoupled
convolutional architecture for the development of a
lightweight model(Howard, 2017). There was two
versions of MobileNet, MobileNet v1 and MobileNet
v2. The updates in MobileNet v2 are the addition
of bottleneck layers and shortcut connections(Sandler
et al., 8 12). Convolutional neural networks have been
used in previous research for face recognition clas-
sification. Thirty-nine (39) classes were included in
the dataset. Fully Connected Layer, pooling layer,
and Convolutional layer without additional architec-
ture were used for training and the accuracy obtained
was 86.71 (Abhirawan et al., 2017).
Cross-Industry Standard Process for Data Min-
ing or CRISP-DM is one of the datamining process
models (datamining framework) which was originally
(1996) built by 5 companies namely Integral Solu-
tions Ltd (ISL), Teradata, Daimler AG, NCR Cor-
poration and OHRA (Mauritsius and Binsar, 2020).
CRISP-DM has the advantage over other models of a
clear definition of the Business Understanding phase.
This phase is not at all considered in detail in other
Data Mining models(Chapman, 2020). Deep learn-
ing has been used in various areas such as computer
vision, natural language processing, audio recogni-
tion, including face recognition. Deep learning is a
multi−layer algorithm for extracting characteristics
and identifying edges such as letters, numbers, faces,
etc. (Farayola and Dureja, 2020).
Convolutional is a subset of deep neural networks
Muttaqin, M., Aziz, A., Irmayanti, D. and Sumanto, .
Convolutional Neural Network Face Recognition for Lecturer Attendance.
DOI: 10.5220/0012447800003848
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 3rd International Conference on Advanced Information Scientific Development (ICAISD 2023), pages 255-261
ISBN: 978-989-758-678-1
Proceedings Copyright © 2024 by SCITEPRESS – Science and Technology Publications, Lda.
255