ANALISIS KOMPARATIF ARSITEKTUR RESNET50 DAN VGG19 DALAM KLASIFIKASI CITRA KAIN ULOS BATAK TOBA

Nova Enjelina Pakpahan, . (2025) ANALISIS KOMPARATIF ARSITEKTUR RESNET50 DAN VGG19 DALAM KLASIFIKASI CITRA KAIN ULOS BATAK TOBA. Skripsi thesis, Universitas Pembangunan Nasional Veteran Jakarta.

[img] Text
ABSTRAK.pdf

Download (235kB)
[img] Text
AWAL.pdf

Download (764kB)
[img] Text
BAB 1.pdf
Restricted to Repository UPNVJ Only

Download (309kB)
[img] Text
BAB 2.pdf
Restricted to Repository UPNVJ Only

Download (1MB)
[img] Text
BAB 3.pdf
Restricted to Repository UPNVJ Only

Download (374kB)
[img] Text
BAB 4.pdf
Restricted to Repository UPNVJ Only

Download (2MB)
[img] Text
BAB 5.pdf

Download (283kB)
[img] Text
DAFTAR PUSTAKA.pdf

Download (242kB)
[img] Text
RIWAYAT HIDUP.pdf
Restricted to Repository UPNVJ Only

Download (157kB)
[img] Text
LAMPIRAN.pdf
Restricted to Repository UPNVJ Only

Download (236kB)
[img] Text
HASIL PLAGIARISME.pdf
Restricted to Repository staff only

Download (19MB)
[img] Text
ARTIKEL KI.pdf
Restricted to Repository staff only

Download (1MB)

Abstract

This study aims to classify the motifs of Ulos Batak Toba fabric using deep learning methods based on pretrained convolutional neural network (CNN) architectures, namely ResNet50 and VGG19. The training and testing processes were conducted on collected Ulos motif images, which were processed through various experiments, including preprocessing techniques such as sharpening and denoising, as well as architectural modifications such as replacing the GlobalAveragePooling layer with GlobalMaxPooling. The study consisted of five experiments combining different preprocessing methods and architectural variations. The results showed that the VGG19 model with GlobalMaxPooling (Experiment 5) achieved the best performance with a test accuracy of 89.58%, while the best-performing ResNet50 model under the same configuration achieved an accuracy of 61.46%. The experiments demonstrated that inconsistent application of preprocessing across training, validation, and test data can reduce the model’s generalization capability. Furthermore, architectural modifications had a greater impact on improving accuracy. The study concludes that both architecture selection and preprocessing significantly affect the classification performance of Ulos motif images. Moreover, larger model architectures do not always yield better results, depending on the suitability to the characteristics of the dataset used.

Item Type: Thesis (Skripsi)
Additional Information: [No.Panggil: 2110511090] [Pembimbing 1: Jayanta] [Pembimbing 2: Nurul Afifah Arifuddin] [Penguji 1: Musthofa Galih Pradana] [Penguji 2: Kharisma Wiati Gusti]
Uncontrolled Keywords: Batak Toba Ulos, image classification, VGG19, ResNet50, deep learning
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Q Science > QA Mathematics > QA76 Computer software
Divisions: Fakultas Ilmu Komputer > Program Studi Informatika (S1)
Depositing User: NOVA ENJELINA PAKPAHAN
Date Deposited: 04 Sep 2025 07:33
Last Modified: 04 Sep 2025 07:33
URI: http://repository.upnvj.ac.id/id/eprint/37804

Actions (login required)

View Item View Item