PERBANDINGAN KINERJA ALGORITMA CONVOLUTIONAL NEURAL NETWORK (CNN) PADA ARSITEKTUR MOBILENETV1 DAN MOBILENETV2 DALAM KLASIFIKASI PAKAIAN BERDASARKAN MUSIM

MASHYTA DIAN ADELIA PUTRI, . (2026) PERBANDINGAN KINERJA ALGORITMA CONVOLUTIONAL NEURAL NETWORK (CNN) PADA ARSITEKTUR MOBILENETV1 DAN MOBILENETV2 DALAM KLASIFIKASI PAKAIAN BERDASARKAN MUSIM. Sarjana thesis, UNIVERSITAS NEGERI JAKARTA.

	Text COVER.pdf Download (667kB)
	Text BAB I.pdf Download (282kB)
	Text BAB II.pdf Restricted to Registered users only Download (718kB) | Request a copy
	Text BAB III.pdf Restricted to Registered users only Download (424kB) | Request a copy
	Text BAB IV.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text BAB V.pdf Restricted to Registered users only Download (274kB) | Request a copy
	Text DAFTAR PUSTAKA.pdf Download (238kB)
	Text LAMPIRAN.pdf Restricted to Registered users only Download (3MB) | Request a copy

Abstract

Klasifikasi pakaian berdasarkan musim pada platform e-commerce membutuhkan model deep learning yang akurat sekaligus efisien secara komputasi, namun kajian perbandingan antar arsitektur ringan masih terbatas. Penelitian ini bertujuan membandingkan kinerja algoritma Convolutional Neural Network (CNN) pada arsitektur MobileNetV1 dan MobileNetV2 dalam klasifikasi pakaian berdasarkan musim (Spring, Summer, Fall, Winter). Penelitian menggunakan dataset Fashion Product Images (Small) dari Kaggle yang terdiri dari 44.441 citra produk, dengan pendekatan Knowledge Discovery in Databases (KDD), strategi transfer learning dan fine tuning, serta 5-Fold Cross Validation pada enam skenario hyperparameter. Kinerja dievaluasi menggunakan metrik akurasi, presisi, recall, dan F1-score. Hasil pada skenario terbaik menunjukkan MobileNetV2 meraih akurasi 84,14% (F1-score 84,09%) dan MobileNetV1 meraih akurasi 84,13% (F1-score 84,17%), dengan selisih yang tidak signifikan secara praktis. Dari sisi efisiensi komputasi, MobileNetV1 lebih unggul dengan waktu pelatihan 3 jam 54 menit dibandingkan MobileNetV2 yang memerlukan 4 jam 58 menit. Penelitian ini menyimpulkan bahwa MobileNetV1 lebih direkomendasikan untuk perangkat dengan sumber daya komputasi terbatas tanpa mengorbankan akurasi secara berarti, dan strategi fine tuning terbukti krusial dalam meningkatkan generalisasi model. Kata Kunci: Convolutional Neural Network, MobileNetV1, MobileNetV2, Klasifikasi Pakaian, Musim, Transfer Learning, Fine Tuning, Deep Learning ***** Clothing classification by season on e-commerce platforms requires a deep learning model that is both accurate and computationally efficient, yet comparative studies across lightweight architectures remain limited. This study aims to compare the performance of Convolutional Neural Network (CNN) algorithms on MobileNetV1 and MobileNetV2 architectures in classifying clothing based on season (Spring, Summer, Fall, Winter). The study utilized the Fashion Product Images (Small) dataset from Kaggle, consisting of 44,441 product images, employing a Knowledge Discovery in Databases (KDD) approach, transfer learning and fine-tuning strategies, and 5-Fold Cross Validation across six hyperparameter scenarios. Performance was evaluated using accuracy, precision, recall, and F1-score metrics. Results from the best-performing scenario show that MobileNetV2 achieved an accuracy of 84.14% (F1-score 84.09%) and MobileNetV1 achieved an accuracy of 84.13% (F1-score 84.17%), with a difference that is not practically significant. In terms of computational efficiency, MobileNetV1 proved superior with a training time of 3 hours 54 minutes compared to MobileNetV2, which required 4 hours 58 minutes. This study concludes that MobileNetV1 is more recommended for devices with limited computational resources without meaningfully sacrificing accuracy, and that fine-tuning strategies are proven to be crucial in improving model generalization. Keywords: Convolutional Neural Network, MobileNetV1, MobileNetV2, Clothing Classification, Season, Transfer Learning, Fine-Tuning, Deep Learning.

Item Type:	Thesis (Sarjana)
Additional Information:	1). Dr. Widodo, S.Kom., M.Kom. ; 2). Neng Ayu Herawati, S.Pd., M.T.
Subjects:	Sains > Matematika > Ilmu Komputer
Divisions:	FT > S1 Pendidikan Teknik Informatika Komputer
Depositing User:	Mashyta Dian Adelia Putri .
Date Deposited:	24 Jul 2026 07:18
Last Modified:	24 Jul 2026 07:18
URI:	http://repository.unj.ac.id/id/eprint/67141

Actions (login required)

View Item