Pengumpulan Dan Klasifikasi Gambar Rumah Adat Indonesia Dengan Metode Deep Learning

Denova, Alfadito Aulia (2025) Pengumpulan Dan Klasifikasi Gambar Rumah Adat Indonesia Dengan Metode Deep Learning. Project Report. [s.n.], [s.l.]. (Unpublished)

Text 5025211157-Project_Report.pdf - Accepted Version Restricted to Repository staff only Download (9MB) | Request a copy

Abstract

Indonesia memiliki keberagaman budaya yang tercermin melalui rumah adat dari setiap provinsi. Di era digital, upaya pelestarian dan dokumentasi visual rumah adat dapat ditunjang melalui pemanfaatan teknologi pengumpulan dan klasifikasi data berbasis kecerdasan buatan. Kerja praktik ini bertujuan untuk mengumpulkan dataset gambar rumah adat Indonesia dalam skala besar menggunakan metode web crawling dan web scraping melalui Google Custom Search Engine (CSE), kemudian melakukan proses penyaringan dan penyuntingan manual untuk memastikan relevansi serta kualitas data. Dataset yang telah diproses digunakan untuk pelatihan model klasifikasi gambar berbasis deep learning menggunakan arsitektur transfer learning ResNet18 dan EfficientNet. Pelatihan model dilakukan pada dataset yang telah melalui proses augmentasi, resize, dan normalisasi piksel. Evaluasi performa menggunakan metrik akurasi, precision, recall, F1-score, confusion matrix, serta analisis loss progression untuk memantau indikasi overfitting. Hasil dari penelitian ini mendapatkan 1199 gambar rumah adat yang mampu digunakan untuk proses training serta mendapatkan model dengan akurasi tertinggi 0.76.
====================================================================================================================================
Indonesia’s cultural diversity is reflected in the traditional houses found across its provinces. In the digital era, efforts to preserve and visually document these traditional houses can be supported through data-driven and AI-based technologies. This internship project aims to collect a large-scale image dataset of Indonesian traditional houses using web crawling and web scraping methods via the Google Custom Search Engine (CSE), followed by manual filtering and editing to ensure data relevance and quality. The processed dataset is then used to train deep learning–based image classification models employing transfer learning architectures, namely ResNet18 and EfficientNet. Model training is conducted on augmented, resized, and normalized images. Performance evaluation includes accuracy, precision, recall, F1-score, confusion matrix, and loss-progression analysis to monitor potential overfitting. The study resulted in a dataset of 1,199 usable images and achieved a highest model accuracy of 0.76.

Item Type:	Monograph (Project Report)
Uncontrolled Keywords:	Rumah Adat, Crawling dan Scraping, ResNet18, EfficientNet, Transfer Learning, Traditional Houses
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence
Divisions:	Faculty of Information Technology > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Alfadito Aulia Denova
Date Deposited:	15 Dec 2025 04:20
Last Modified:	15 Dec 2025 04:20
URI:	http://repository.its.ac.id/id/eprint/128950

Actions (login required)

View Item