PENERAPAN METODE CONVOLUTIONAL NEURAL NETWORK DENGAN ARSITEKTUR RESNET-50 UNTUK PENGENALAN EKSPRESI WAJAH

SALASTA, RIFKY IZHA (2024) PENERAPAN METODE CONVOLUTIONAL NEURAL NETWORK DENGAN ARSITEKTUR RESNET-50 UNTUK PENGENALAN EKSPRESI WAJAH. Other thesis, UPN Veteran Yogyakarta.

	Text ABSTRAK.pdf Download (176kB)
	Text COVER.pdf Download (158kB)
	Text DAFTAR ISI.pdf Download (169kB)
	Text DAFTAR PUSTAKA.pdf Download (160kB)
	Text LEMBAR PENGESAHAN PEMBIMBING.pdf Download (744kB)
	Text LEMBAR PENGESAHAN PENGUJI.pdf Download (751kB)
	Text SKRIPSI FULL_RIFKY IZHA SALASTA_123170032.pdf Restricted to Repository staff only Download (4MB)

Abstract

viii
ABSTRAK
Pengenalan ekspresi wajah merupakan suatu teknik untuk memahami emosi manusia
dari ekspresi yang ditunjukkan sebagai reaksi terhadap sesuatu yang terjadi dari lingkungan.
Di dunia digital ini, pengenalan ekspresi wajah dapat diterapkan secara luas. Misalnya, dapat
digunakan untuk memahami ekspresi manusia dalam video conference online. Dalam video
meeting online, ada aspek interaksi mikro yang hilang jika dibandingkan dengan interaksi
sosial langsung. Pengenalan ekspresi wajah dalam videomeeting online diharapkan dapat
meningkatkan pemahaman interaksi pengguna
Penelitian ini dilakukan dengan metode Convolutional Neural Network (CNN)
arsitektur Resnet50. Pemilihan metode tersebut dilakukan dengan alasan penggunaan
Residual Network yang mengurangi kompleksitas dan memecahkan masalah degradasi
sambil tetap mempertahankan kinerja yang baik.
Hasil dari pengujian pengenalan ekspresi wajah dengan metode CNN arsitektur
Resnet50 didapatkan hasil akurasi pelatihan data sebesar 94,83% dan akurasi validasi
sebesar 81,02%. Kemudian dari pengujian yang dilakukan dengan metode confusion matrix
didapatkan nilai presisi sebesar 82,12%, recall 78,17%, f1-score 82,27% dan memiliki nilai
akurasi sebesar 81,5%. Model pengenalan ekspresi wajah yang digunakan dapat
mengklasifikasikan ekspresi wajah dengan cukup baik namun masih memiliki beberapa
kekurangan seperti kesulitan mengenali ekspresi wajah dengan input gambar berwarna
karena pada saat proses pelatihan data dilakukan dengan dataset yang hanya berwarna hitam
putih.
Kata Kunci : Convolutional Neural Network (CNN), Ekspresi Wajah, Facial Expression
Recognition (FER), Deep Leaning, Transfer Learning, Resnet50, Klasifikasi.
ix
ABSTRACT
Facial expression recognition is a technique used to understand human emotions
based on the expressions shown in response to environmental stimuli. In the digital world,
facial expression recognition can be widely applied. For instance, it can be used to
understand human expressions during online video conferences. In online video meetings,
there is a loss of micro-interaction aspects compared to direct social interactions. Facial
expression recognition in online video meetings is expected to enhance user interaction
understanding.
This study was conducted using the Convolutional Neural Network (CNN) method
with the ResNet-50 architecture. This method was chosen due to the use of Residual
Networks, which reduce complexity and solve degradation problems while maintaining
good performance.
The results of facial expression recognition testing using the CNN method with the
ResNet-50 architecture showed a training accuracy of 94.83% and a validation accuracy of
81.02%. Additionally, testing using the confusion matrix method yielded a precision of
82.12%, a recall of 78.17%, an F1-score of 82.27%, and an overall accuracy of 81.5%. The
facial expression recognition model used can classify facial expressions fairly well, but it
still has some shortcomings, such as difficulty recognizing facial expressions from colored
images, as the training data was processed using a black-and-white dataset.
Keywords : Convolutional Neural Network (CNN), Facial Expression, Facial Expression
Recognition (FER), Deep Leaning, Transfer Learning, Resnet50, Classification.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Convolutional Neural Network (CNN), Facial Expression, Facial Expression Recognition (FER), Deep Leaning, Transfer Learning, Resnet50, Classification.
Subjects:	Z Bibliography. Library Science. Information Resources > ZA Information resources
Divisions:	Faculty of Engineering, Science and Mathematics > School of Engineering Sciences
Depositing User:	Eko Yuli
Date Deposited:	22 Jul 2024 02:15
Last Modified:	22 Jul 2024 02:15
URI:	http://eprints.upnyk.ac.id/id/eprint/40323

Actions (login required)

View Item