Putra, I Nyoman Tri Anindia (2026) Rekonstruksi dan Penyelesaian Objek 3D Sebagai Upaya Pelestarian Budaya Ukiran Khas Bali. Doctoral thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 7025231008-Doctoral.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
7025231008-Doctoral.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy |
Abstract
Dalam konteks keanekaragaman dan kekayaan budaya Bali, ukiran kayu tradisional menonjol sebagai salah satu warisan yang memukau dengan detail rumit dan sarat makna. Namun, keaslian dan kelangsungan budaya ini terancam oleh arus globalisasi dan modernisasi. Inisiatif awal untuk melestarikan budaya ini dilakukan melalui pendekatan digitalisasi menggunakan teknologi fotogrametri dan perangkat lunak seperti Polycam dan Blender pada tahun 2023. Proses ini memungkinkan pembuatan arsip digital objek-objek kultural, seperti topeng dan struktur arsitektural rumah tulang Bali, yang tidak hanya menawarkan reproduksi visual tetapi juga platform interaktif untuk eksplorasi budaya.
Inisiatif ini menghadapi tantangan teknis dan logistik yang signifikan, termasuk tingginya kebutuhan waktu dan investasi finansial untuk menghasilkan model 3D yang akurat. Untuk mengatasi hal tersebut, penelitian ini mengeksplorasi teknologi 3D reconstruction dan 3D completion secara ekstensif dengan membandingkan lima arsitektur Deep Learning, yaitu AutoSDF, SDFusion, TripoSR mandiri, serta pendekatan hibrida SDFusion + TripoSR dan SF3D + TripoSR. Penelitian ini membangun dataset BaliCarve3D dan BaliMask3D menggunakan metode fotogrametri sebagai data primer untuk evaluasi. Berdasarkan hasil evaluasi kuantitatif yang komprehensif, kombinasi metode SF3D + TripoSR terbukti memberikan performa rekonstruksi terbaik. Pendekatan ini mencapai akurasi struktural dan fidelitas Topologi yang luar biasa dengan rata-rata metrik evaluasi puncak: F1 Score mencapai 0,8885, Intersection over Union (IoU) sebesar 0,8016, serta tingkat kesalahan yang sangat rendah pada Chamfer Distance (CD) 0,0266 dan Uniform Hausdorff Distance (UHD) 0,0410, Model ini juga menunjukkan konsistensi Topological Morphological Distance (TMD) yang sangat baik hingga mencapai nilai absolut 0,0140, yang secara efektif mempertahankan ornamen detail dan topologi skeleton ruang objek bahkan pada struktur yang kompleks. Hasil penelitian ini menyediakan fondasi teknologi state-of-the-art serta dataset berharga untuk pelestarian digital, membuka peluang revolusioner bagi generasi masa depan untuk mendokumentasikan dan memahami kekayaan budaya Bali secara utuh.
===================================================================================================================================
Within the context of Bali's rich and diverse culture, traditional wood carvings stand out as a magnificent cultural heritage, characterized by their intricate details and profound meanings. However, the authenticity and continuity of this culture are increasingly threatened by globalization and modernization. An initial initiative to preserve this heritage was conducted through a digitization approach using photogrammetry technologies and software such as Polycam and Blender in 2023. This process enabled the creation of digital archives of cultural objects, such as masks and the architectural structures of traditional Balinese houses, offering not only visual reproduction but also an interactive platform for cultural exploration. This initiative encountered significant technical and logistical challenges, including the substantial time and financial investments required to produce accurate 3D models. To overcome these limitations, this research extensively explores 3D reconstruction and 3D completion technologies by comparing five Deep Learning architectures: standalone AutoSDF, SDFusion, and TripoSR, as well as the hybrid approaches of SDFusion + TripoSR and SF3D + TripoSR. The study developed the BaliCarve3D and BaliMask3D datasets using photogrammetry methods as primary data for evaluation. Based on comprehensive quantitative evaluations, the hybrid combination of SF3D + TripoSR proved to yield the best reconstruction performance. This approach achieved exceptional structural accuracy and topological fidelity with peak average evaluation metrics: an F1 Score reaching 0,8885, an Intersection over Union (IoU) of 0,8016, and exceptionally low error rates with a Chamfer Distance (CD) of 0,0266 and a Uniform Hausdorff Distance (UHD) of 0,0410, The model also demonstrated excellent Topological Morphological Distance (TMD) consistency, reaching an absolute value of 0,0140, effectively preserving the detailed ornaments and spatial skeleton topology of the objects even in complex structures. The results of this research provide a state-of-the-art technological foundation and a valuable dataset for digital preservation, opening revolutionary opportunities for future generations to comprehensively document and understand the wealth of Balinese culture.
| Item Type: | Thesis (Doctoral) |
|---|---|
| Uncontrolled Keywords: | 3D Completion, 3D Reconstruction, Computer Vision, Deep Learning, Digital Heritage |
| Subjects: | G Geography. Anthropology. Recreation > GF Human ecology. Anthropogeography > GF78 T73 Sustainable living. Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > Q Science (General) > Q337.5 Pattern recognition systems T Technology > T Technology (General) > T57.5 Data Processing T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing. |
| Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55001-(S3) PhD Thesis (Comp Science) |
| Depositing User: | I Nyoman Tri Anindia Putra |
| Date Deposited: | 22 Jul 2026 00:39 |
| Last Modified: | 22 Jul 2026 00:39 |
| URI: | http://repository.its.ac.id/id/eprint/135026 |
Actions (login required)
 |
View Item |