Klasifikasi Citra Fungi Mikroskopis Menggunakan VGG16 dengan Mekanisme Attention

Fauzi, Muhammad Zulfikar (2025) Klasifikasi Citra Fungi Mikroskopis Menggunakan VGG16 dengan Mekanisme Attention. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6025221060-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB)

Abstract

Fungi memainkan peran penting dalam ekosistem, berfungsi sebagai pengurai dan kontributor siklus nutrisi. Meskipun terdapat fungi yang tidak berbahaya dan berkontribusi positif terhadap lingkungan, spesies fungi tertentu dapat menimbulkan risiko bagi kesehatan manusia dan makhluk hidup lainnya. Identifikasi fungi mikroskopis berkontribusi terhadap pelestarian lingkungan dan kesehatan manusia. Salah satu metode klasifikasi fungi adalah dengan menggunakan teknologi deep learning. Penelitian ini melakukan klasifikasi terhadap gambar fungi mikroskopis menggunakan pendekatan deep learning. Dataset yang digunakan dalam penelitian ini adalah dataset DeFungi, yang berisi gambar dari lima macam spesies fungi. Arsitektur yang digunakan adalah VGG16 dengan modifikasi tambahan attention mechanism, seperti penambahan Convolutional Block Attention Module (CBAM) diantara blok-blok konvolusi. Teknik augmentasi data diterapkan untuk meningkatkan variasi data pelatihan dengan menghasilkan data sintetis melalui transformasi seperti flipping dan rotation. Selain itu, teknik image filtering digunakan untuk menonjolkan fitur visual pada gambar, sehingga dapat meningkatkan kemampuan model dalam melakukan klasifikasi. Dataset dibagi menjadi data latih, data validasi, dan data uji dengan rasio 65:15:20. Performa model dievaluasi menggunakan metrik akurasi dan confusion matrix untuk mengukur keberhasilan klasifikasi. Penambahan attention mechanism pada VGG16 relatif meningkatkan performa, akan tetapi juga menimbulkan overfitting. Problematika ini berhasil diatasi menggunakan arsitektur modifikasi VGG16 yang diajukan pada penelitian ini. Penggunaan teknik image filtering juga meningkatkan performa model lebih lanjut. Didapatkan akurasi paling tinggi bernilai 94,33% dengan penggunaan image filtering High Frequency Emphasis (HFE).
==================================================================================================================================
Fungi play an essential role in ecosystems, functioning as decomposers and contributors to nutrient cycles. While many fungi are harmless and contribute positively to the environment, certain fungal species can pose risks to human health and other living organisms. The identification of microscopic fungi contributes to environmental preservation and human health. One method of classifying fungi is through the use of deep learning technology. This study classifies images of microscopic fungi using a deep learning approach. The dataset used in this study is the DeFungi dataset, which contains images of five different fungal species. The architecture employed is VGG16, with additional modifications in the form of an attention mechanism, specifically the addition of a Convolutional Block Attention Module (CBAM) between convolutional blocks. Data augmentation techniques were applied to increase the diversity of training data by generating synthetic data through transformations such as flipping and rotation. Additionally, image filtering techniques were used to highlight visual features in the images, improving the model’s classification capability. The dataset was divided into training, validation, and testing sets with a ratio of 65:15:20. Model performance was evaluated using accuracy metrics and a confusion matrix to measure classification success. The addition of the attention mechanism to VGG16 relatively improved performance but also caused overfitting. This issue was successfully addressed using the modified VGG16 architecture proposed in this study. The use of image filtering techniques further enhanced model performance, achieving the highest accuracy of 94.33% with the application of the High Frequency Emphasis (HFE) image filtering technique.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Attention Mechanism, Deep Learning, Fungi, Image Classification, VGG16
Subjects:	Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Muhammad Zulfikar Fauzi
Date Deposited:	29 Jan 2026 04:14
Last Modified:	29 Jan 2026 04:14
URI:	http://repository.its.ac.id/id/eprint/130785

Available Versions of this Item

• Klasifikasi Citra Fungi Mikroskopis Menggunakan VGG16 dengan Mekanisme Attention. (deposited 29 Jan 2026 05:24)
  • Klasifikasi Citra Fungi Mikroskopis Menggunakan VGG16 dengan Mekanisme Attention. (deposited 29 Jan 2026 04:14) [Currently Displayed]

Actions (login required)

View Item