Klasifikasi Tulisan Tangan Aksara Sunda Dengan Model Inception-ResNetV2 dan Metode Transfer Learning

Paramitha, Clara Diva (2025) Klasifikasi Tulisan Tangan Aksara Sunda Dengan Model Inception-ResNetV2 dan Metode Transfer Learning. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 20081010097_Cover.pdf Download (2MB)
	Text (Bab 1) 20081010097_Bab 1.pdf Download (160kB)
	Text (Bab 2) 20081010097_Bab 2.pdf Restricted to Repository staff only until 15 September 2027. Download (1MB)
	Text (Bab 3) 20081010097_Bab 3.pdf Restricted to Repository staff only until 15 September 2027. Download (1MB)
	Text (Bab 4) 20081010097_Bab 4.pdf Restricted to Repository staff only until 15 September 2027. Download (2MB)
	Text (Bab 5) 20081010097_Bab 5.pdf Download (135kB)
	Text (Daftar Pustaka) 20081010097_Daftar Pustaka.pdf Download (145kB)
	Text (Lampiran) 20081010097_Lampiran.pdf Restricted to Repository staff only until 15 September 2027. Download (3MB)

Abstract

Indonesia is a widely known archipelago state, comprising over 17.000 islands, sparcely spread among 34 provinces, based on the records accumulated by Badan Pusat Statistik (BPS) per 2022. This is one of Indonesia’s unique appeals, alluring many nations to visit and fortuitously, bringing their own diverse cultures—one of them being Hindu-Buddhist culture. As a result, an inevitable assimilation occured between local and Hindu-Buddhist cultures, bringing new knowledge that significantly changed Indonesia’s civilization, particularly in the field of literature. Gradually, Old Pallava script evolved into Kawi script, and over time, Javanese Kawi script was adapted into various regional scripts, such as Javanese, Balinese, Sundanese, and many more. The same way Javanese has its own script, Sundanese also has one. These days, the use of Sundanese scripts has substantially declined, resulting in a lack of awareness about its existence in nowadays society. Moreover, during its usage, it’s more common to write the scripts by hands, which led to many variations of form. These factors have caused growing confusion—or worse, misunderstanding—within society, rooted in illegible handwriting or certain letters looking too similar to another. Stemming from these issues, in order to save time and human resources in detecting which letter is which, it requires a reliable help for automatic detection and classification. Enabled by technology’s rapid evolution, the field of Deep Learning is looking remarkably reliable to provide for solutions to these issues. This research implemented Transfer Learning method and setting Inception-ResnetV2—a combination of two CNN famous architectures—as its base model to be tested against varying optimizers and learning rates using two types of Transfer Learning approaches and two types of epoch. The data used in this research was partly obtained from Github and partly from self-collection amounted to 18 classes of Ngalagena Sundanese script and 5 classes of borrowed Sundanese script. The outcome of this research revealed that a combination consisted of Inception-ResnetV2 architecture, fine-tuning based model, four optimizers 20 epoch proved to be adept at a classifying task. The results varied depending severely on the learning rate used. Keywords: Sundanese scripts, Deep Learning, CNN, Inception-ResnetV2, Transfer Learning

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Junaidi, Achmad	NIDN0710117803	achmadjunaidi.if@upnjatim.ac.id
Thesis advisor	Al Haromainy, Muhammad Muharrom	NIDN0701069503	muhammad.muharrom.if@upnjatim.ac.id

Subjects:

T Technology > T Technology (General)

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Clara Diva Paramitha

Date Deposited:

15 Sep 2025 06:21

Last Modified:

15 Sep 2025 06:21

URI:

https://repository.upnjatim.ac.id/id/eprint/43524