IMPLEMENTASI CNN DENGAN TRANSFER LEARNING UNTUK KLASIFIKASI TUTUPAN HUTAN MENGGUNAKAN CITRA SENTINEL-2 DI KAWASAN PEGUNUNGAN ARJUNO–WELIRANG

Yusuf, M Saaduddin Abdillah (2026) IMPLEMENTASI CNN DENGAN TRANSFER LEARNING UNTUK KLASIFIKASI TUTUPAN HUTAN MENGGUNAKAN CITRA SENTINEL-2 DI KAWASAN PEGUNUNGAN ARJUNO–WELIRANG. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) cover.pdf Download (781kB)
	Text (Bab 1) BAB 1.pdf Download (205kB)
	Text (Bab 2) BAB 2.pdf Restricted to Repository staff only until 29 September 2029. Download (481kB)
	Text (Bab 3) BAB 3.pdf Restricted to Repository staff only until 29 September 2029. Download (598kB)
	Text (Bab 4) BAB 4.pdf Restricted to Repository staff only until 29 September 2029. Download (2MB)
	Text (Bab 5) BAB 5.pdf Download (184kB)
	Text (Daftar Pustaka) Daftar Pustaka.pdf Download (200kB)
	Text (Lampiran) Lampiran.pdf Restricted to Repository staff only until 29 September 2029. Download (761kB)

Abstract

Anthropogenic land conversion in the Arjuno–Welirang Mountain Area creates a need for more objective methods of forest cover monitoring. This study aims to compare the performance of three Convolutional Neural Network (CNN) architectures, namely MobileNetV2, ResNet-50, and VGG-16, using Transfer Learning for forest cover classification. The dataset consists of single-temporal composite Sentinel-2 imagery from 2025, divided into 224 × 224 pixel patches and grouped into three classes: Forest, Natural Non-Forest, and Man-made Non-Forest based on SNI 7645:2010. The experiments compared Feature Extraction and Fine-Tuning strategies for each architecture. The results on the sterile test dataset showed that Partial Fine-Tuning improved performance compared with Feature Extraction across the three architectures. The best-performing model was MobileNetV2 with Partial Fine-Tuning (S2), achieving an Overall Accuracy of 95.80% and a Macro F1-Score of 96.05%. For the Man-made Non-Forest class, the model achieved Precision, Recall, and F1-Score of 96.30%. The S2 model had a model file size of 24.12 MB and an accuracy gap of 1.51% between the training and test data. The model was then integrated into a Flask-based ForestGuard web prototype to display classification results as a grid map and support area risk status determination based on the classified land cover.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Wati, Seftin Fitri Ana	NIDN21219910320267	seftin.fitri.si@upnjatim.ac.id
Thesis advisor	Pahlawan, Muhammad Reza	NIDN199805162025061005	muhammad_reza.si@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Q Science > QA Mathematics > QA76.6 Computer Programming
Q Science > QA Mathematics > QA76.87 Neural computers
T Technology > T Technology (General)

Divisions:

Faculty of Computer Science > Departemen of Information Systems

Depositing User:

Sa'aduddin Abdillah Yusuf

Date Deposited:

29 Sep 2026 07:51

Last Modified:

29 Sep 2026 07:51

URI:

https://repository.upnjatim.ac.id/id/eprint/60280