Metode Classical-Quantum Neural Network Berbasis Transfer Learning Untuk Klasifikasi Citra Daun Padi Berpenyakit

Muntazhar, Ahmad Zaki Al (2025) Metode Classical-Quantum Neural Network Berbasis Transfer Learning Untuk Klasifikasi Citra Daun Padi Berpenyakit. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6002231014-Master_Thesis.pdf Restricted to Repository staff only until 1 April 2027. Download (20MB) | Request a copy

Abstract

Indonesia adalah negara agraris yang sebagian besar masyarakatnya bekerja di sektor pertanian. Produksi padi Indonesia mengalami penurunan dari tahun ke tahun hingga menyebabkan terjadinya inflasi harga beras. Sehinggga untuk meningkatkan kuantitas produksi padi Indonesia identifikasi dini penyakit pada daun padi menjadi penting dilakukan. Metode klasifikasi penyakit tradisional sering kali mengalami kesulitan dalam hal akurasi dan waktu pemrosesan, terutama ketika berhadapan dengan dataset skala besar. Untuk mengatasi tantangan ini, penelitian ini mengonstruksi penggabungan komputasi klasik dan kuantum. Secara khusus, metode Classical-Quantum Neural Network (CQNN) dikembangkan dengan memanfaatkan teknik transfer learning (TL) dan quantum layer. Teknik TL pada penelitian tesis ini memanfaatkan tiga jenis pretrained model yaitu MobileNet, ResNet dan EfficientNet. Sedangkan quantum layer gerbang Rot dikonstruksi dengan jumlah 8 qubit. Berdasarkan hasil evaluasi pelatihan, model CQNN dengan EfficientNet dan optimizer Adam mendapatkan akurasi validasi sebesar 97,46%. Model tersebut mendapatkan nilai akurasi terendah pada skema kekaburan σ = 2 sebesar 84,68%. Sedangkan akurasi terendah dengan nilai 91,33% terjadi pada skema kecerahan dengan pengurangan 100 intensitas cahaya.
=================================================================================================================================
Indonesia is an agrarian country where most of the people work in the agricultural sector. Indonesia’s rice production has decreased from year to year, causing rice price inflation. Therefore, to increase the quantity of rice production in Indonesia, early identification of diseases on rice leaves is important. Traditional disease classification methods often have difficulties in terms of accuracy and processing time, especially when dealing with largescale datasets. To overcome this challenge, this research constructs a fusion of classical and quantum computing. Specifically, the Classical-Quantum Neural Network (CQNN) method is developed by utilizing the transfer learning (TL) and quantum layer techniques. The TL technique in this thesis research utilizes three types of pre-trained models namely MobileNet, ResNet and EfficientNet. While the Rot gate quantum layer is constructed with 8 qubits. Based on the training evaluation results, the CQNN model with EfficientNet and Adam’s optimizer obtained a validation accuracy of 97.46%. The model gets the lowest accuracy value on the σ = 2 blurring scheme of 84.68%. While the lowest accuracy with a value of 91.33% occurred in the brightness scheme with a reduction of 100 light intensities.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Quantum neural network, transfer learning, quantum machine learning, rice leaf disease classification, precision agriculture. Quantum neural network, pembelajaran transfer, quantum machine learning, klasifikasi citra daun padi berpenyakit, presisi agrikultur
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.6 Computer programming. Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QA Mathematics > QA76.9 Computer algorithms. Virtual Reality. Computer simulation. S Agriculture > SB Plant culture > SB191.R5 Rice farming
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44101-(S2) Master Thesis
Depositing User:	Ahmad Zaki Al Muntazhar
Date Deposited:	07 Feb 2025 03:02
Last Modified:	07 Feb 2025 03:02
URI:	http://repository.its.ac.id/id/eprint/117814

Actions (login required)

View Item