Klasifikasi Citra Penyakit Daun Padi Menggunakan Arsitektur Convolutional Neural Network

Rizal, Mohammad Hasan Tajuk (2025) Klasifikasi Citra Penyakit Daun Padi Menggunakan Arsitektur Convolutional Neural Network. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 21082010138_cover.pdf Download (1MB)
	Text (Bab 1) 21082010138_bab1.pdf Download (42kB)
	Text (Bab 2) 21082010138_bab2.pdf Restricted to Repository staff only until 14 November 2028. Download (734kB)
	Text (Bab 3) 21082010138_bab3.pdf Restricted to Repository staff only until 14 November 2028. Download (271kB)
	Text (Bab 4) 21082010138_bab4.pdf Restricted to Repository staff only until 14 November 2028. Download (1MB)
	Text (Bab 5) 21082010138_bab5.pdf Download (12kB)
	Text (Daftar pustaka) 21082010138_daftarpustaka.pdf Download (266kB)
	Text (Lampiran) 21082010138_lampiran.pdf Restricted to Repository staff only until 14 November 2028. Download (394kB)

Abstract

Rice is a primary food commodity in Indonesia, yet its production is often threatened by various leaf diseases that can lead to crop failure. Early and accurate detection is key to effective management. These diseases often show distinct visual symptoms in the form of spots with specific patterns and colors on the leaves. Rice leaves serve as an ideal identification object due to their wide surface area, making these changes easy to observe and suitable for accurate analysis using image processing technology. This research aims to build an image classification system for rice leaf diseases using a Convolutional Neural Network (CNN) by comparing three pre-trained architectures: MobileNetV3, EfficientNet-B0, and ShuffleNetV2. The methodology employed is transfer learning followed by a staged fine-tuning process to adapt the models from a public dataset to a target dataset with a different domain. Nine modeling scenarios with varying hyperparameters, such as learning rate and dropout, were tested to find the best configuration. The evaluation results show that the fine-tuning process significantly improved the performance of the MobileNetV3 and EfficientNet-B0 models. The best model was achieved by Scenario 6 (EfficientNet-B0), which attained the highest F1-Score of 0.70. Conversely, the ShuffleNetV2 architecture proved less capable of adapting to the target domain data. The best-performing model was then successfully converted to the TensorFlow Lite (TFLite) format and implemented in a functional Android application named "PadiCare." This study concludes that the EfficientNet-B0 architecture is the most effective choice for this task and has the potential to be implemented as a practical support tool for farmers.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Arifiyanti, Amalia Anjani	NIDN0712089201	amalia_anjani.fik@upnjatim.ac.id
Thesis advisor	Sembilu, Nambi	NIDN9990610124	nambi.si@upnjatim.ac.id

Subjects:

Q Science > Q Science (General)
T Technology > T Technology (General)

Divisions:

Faculty of Computer Science > Departemen of Information Systems

Depositing User:

Mohammad Hasan Tajuk Rizal

Date Deposited:

11 Feb 2026 07:26

Last Modified:

11 Feb 2026 07:43

URI:

https://repository.upnjatim.ac.id/id/eprint/49341