Pengenalan Wajah Karyawan Menggunakan Algoritma Convolutional Neural Network untuk Sistem Presensi pada CV. Dwi Santoso

Putra, Arya Rizky Tri (2026) Pengenalan Wajah Karyawan Menggunakan Algoritma Convolutional Neural Network untuk Sistem Presensi pada CV. Dwi Santoso. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 22082010067-cover.pdf Download (704kB)
	Text (Bab 1) 22082010067-bab1.pdf Download (342kB)
	Text (Bab 2) 22082010067-bab2.pdf Restricted to Repository staff only until 17 July 2028. Download (723kB)
	Text (Bab 3) 22082010067-bab3.pdf Restricted to Repository staff only until 19 July 2028. Download (504kB)
	Text (Bab 4) 22082010067-bab4.pdf Restricted to Repository staff only until 19 July 2028. Download (2MB)
	Text (Bab 5) 22082010067-bab5.pdf Download (238kB)
	Text (Daftar pustaka) 22082010067-daftarpustaka.pdf Download (195kB)
	Text (Lampiran) 22082010067-lampiran.pdf Restricted to Repository staff only Download (1MB)

Abstract

Manual attendance at CV. Dwi Santoso has several limitations, including data manipulation risk, recording errors, delayed recapitulation, and limited flexibility for mobile, freelance, and outsourced employees. This thesis aims to develop a webbased employee attendance system with face recognition validation using a Convolutional Neural Network based on the MobileNetV2 architecture. The methodology follows CRISP-DM, consisting of Business Understanding, Data Understanding, Data Preparation, Modeling, Evaluation, and Deployment. The dataset consists of facial images from 13 employees captured from front, right, and left angles, producing 650 initial images. Data preparation includes face extraction, image resizing, pixel normalization, rotation augmentation, horizontal flipping, lighting adjustment, grayscale conversion, label encoding, and data splitting using K-Fold Cross Validation. The model was evaluated through five hyperparameter tuning scenarios by varying dropout values and epoch numbers. Evaluation was conducted using validation data and hold-out test data with controlled variations to assess the model robustness under more diverse image conditions. The best result was achieved by Scenario 4 with 0.3 dropout and a maximum of 50 epochs, producing a mean validation accuracy of 0.9923, mean validation loss of 0.0643, test accuracy of 0.9923, test loss of 0.0377, and average precision, recall, and F1-score of 0.99. The best model was implemented into a web-based system using Flask and MySQL to support authentication, employee management, face validation, clock-in, clock-out, and attendance recapitulation.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Arifiyanti, Amalia Anjani	NIDN19920812 2018032 001	amalia_anjani.fik@upnjatim.ac.id
Thesis advisor	Sembilu, Nambi	NIDN199005162024061003	nambi.si@upnjatim.ac.id

Subjects:

T Technology > T Technology (General)

Divisions:

Faculty of Computer Science > Departemen of Information Systems

Depositing User:

Arya Rizky Tri Putra

Date Deposited:

21 Jul 2026 02:44

Last Modified:

21 Jul 2026 02:44

URI:

https://repository.upnjatim.ac.id/id/eprint/56284