| id |
ascaad2023_096 |
| authors |
Şenkal, Habibe; Alaçam, Sema; Güzelci, Orkan; Bayram, Asena; Neftçi, Aras |
| year |
2023 |
| title |
A Deep Learning-Based Model to Estimate Architectural Elements of Classical Ottoman Period Mosques |
| source |
C+++: Computation, Culture, and Context Proceedings of the 11th International Conference of the Arab Society for Computation in Architecture, Art and Design (ASCAAD), University of Petra, Amman, Jordan [Hybrid Conference] 7-9 November 2023, pp. 585-598. |
| summary |
Deep learning algorithms are widely used in architecture for a variety of purposes such as detecting, analyzing, and classifying building types and architectural elements. This paper particularly focuses on the application of deep learning algorithms in the field of architectural heritage. The aim of this paper is to develop a deep learning-based model for predicting and classifying architectural elements, using Süleymaniye Mosque, a masterpiece by Architect Sinan, as a case study. YOLOv4, a CNN-based object recognition algorithm, is employed to identify three distinctive elements of Süleymaniye Mosque: domes, pendentives, and windows. The process of model development includes data collection and refinement, training and testing, as well as validation through architectural element estimation. The F1 score metric is utilized to objectively evaluate the model's performance. The results indicate that the best F1 score for domes (0.80) is achieved in the 6000th iteration, while for windows, the highest F1 score (0.69) is observed in the 3200th iteration. Regarding the pendentive element, both the 650th and 6000th iterations yield a similar F1 score (0.86) as the highest. This study demonstrates the ability of deep learning to recognize historical building elements that belong to a particular style or era. |
| series |
ASCAAD |
| email |
senkal21@itu.edu.tr |
| full text |
 file.pdf (684,032 bytes) |
| references |
Content-type: text/plain
|
| last changed |
2024/02/13 14:41 |
|
