Please use this identifier to cite or link to this item: https://repositori.uma.ac.id/handle/123456789/30266
Title: Klasifikasi Tingkat Kematangan Buah Pisang Berbasis Citra Menggunakan Arsitektur MobileNetV3
Other Titles: Image-Based Banana Ripeness Classification Using MobileNetV3 Architecture
Authors: Octiani, Siska
metadata.dc.contributor.advisor: Sembiring, Arnes
Keywords: klasifikasi citra;kematangan pisang;MobileNetV3;deep learning;aplikasi Android;image classification;banana ripeness;Android application
Issue Date: Feb-2026
Publisher: Universitas Medan Area
Series/Report no.: NPM;228160047
Abstract: Penelitian ini bertujuan mengembangkan dan menerapkan sistem klasifikasi tingkat kematangan buah pisang berbasis citra digital dengan menggunakan arsitektur MobileNetV3 pada aplikasi mobile Android. Fokus masalah terletak pada keterbatasan penilaian manual yang bersifat subjektif, inkonsisten, dan tidak efisien untuk lingkungan industri saat ini. Untuk mengatasi masalah tersebut, acuan teori diambil dari bidang computer vision dan deep learning, khususnya melalui Convolutional Neural Network serta konsep transfer learning. Data diperoleh dengan mengambil citra langsung buah pisang Barangan di Desa Sibolangit, Kabupaten Karo, sebanyak 1.500 gambar yang dibagi ke dalam tiga kategori: mentah, setengah matang, dan matang. Selanjutnya, data diproses melalui tahap prapemrosesan yang mencakup resizing, normalisasi, dan augmentasi data. Model MobileNetV3-Small dan MobileNetV3-Large dilatih menggunakan pendekatan transfer learning dengan parameter pelatihan yang serupa, dan hasilnya dianalisis secara kuantitatif menggunakan metrik seperti akurasi, presisi, recall, F1-score, serta confusion matrix. Kajian ini menemukan bahwa kedua model menampilkan performa yang tinggi, namun MobileNetV3-Small lebih unggul dalam efisiensi komputasi dan waktu inferensi, menjadikannya pilihan yang lebih tepat untuk diimplementasikan di perangkat mobile berdaya rendah dalam sistem klasifikasi kematangan pisang secara real-time. This study aims to develop and implement a digital image-based banana ripeness classification system using the MobileNetV3 architecture in an Android mobile application. The problem focuses on the limitations of manual assessment, which is subjective, inconsistent, and inefficient for current industrial environments. To address this issue, theoretical references are drawn from the fields of computer vision and deep learning, particularly through Convolutional Neural Networks and the concept of transfer learning. The data were obtained by directly capturing images of Barangan bananas in Sibolangit Village, Karo Regency, totaling 1,500 images divided into three categories: unripe, half-ripe, and ripe. The data were then processed through preprocessing stages including resizing, normalization, and data augmentation. The MobileNetV3-Small and MobileNetV3-Large models were trained using a transfer learning approach with similar training parameters, and the results were analyzed quantitatively using metrics such as accuracy, precision, recall, F1-score, and confusion matrix. This study finds that both models demonstrate high performance; however, MobileNetV3-Small outperforms in terms of computational efficiency and inference time, making it a more appropriate choice for implementation on low-resource mobile devices in real-time banana ripeness classification systems.
Description: 53 Halaman
URI: https://repositori.uma.ac.id/handle/123456789/30266
Appears in Collections:SP - Informatic Engineering

Files in This Item:
File Description SizeFormat 
228160047 - Siska Octiani - Fulltext.pdfCover, Abstract, Chapter I, II, III, V, Bibliography2.07 MBAdobe PDFView/Open
228160047 - Siska Octiani - Chapter IV.pdf
  Restricted Access
Chapter IV763.32 kBAdobe PDFView/Open Request a copy


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.