Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Architecture Heritage Recognition Using YOLACT Instance Segmentation
Bahrain has a rich architectural heritage that is reflected in its forts and houses. The forts and houses of Bahrain are not only important cultural landmarks but also serve as examples of the region's architectural styles. These forts and houses are built using locally-sourced materials, such as coral stone and palm wood, and feature distinctive design elements, such as wind towers, courtyards, and arched windows. Even, they serve as important cultural landmarks and provide valuable insights into the region's building techniques. However, the previous methods that are currently employed for architecture form come within the category of subjective analysis. In order to identify architecture form, identify regional architectural style, and provide a quantitative evaluation, quantitative approaches are essential. In this research study, a novel quantitative features based YOLACT object detection model has been employed. Firstly, the Bahrain architectural heritage forts and houses images have been collected from secondary sources with the ramparts, bastions, gateways, watch towers, large windows, high ceilings, wide front porches feature. Secondly, an image resizing and normalization techniques were used to preprocess the images. The preprocessed images have been used to train the object detection model that indentify the eight forts and houses features. Moreover, the YOLACT model obtained the 599 features of forts and 390 features of houses from preprocessed images directly with several epochs and learning rate. Furthermore, the performance of YOLACT model has been analyzed through precision, recall, F1-score, AP parameters. YOLACT model exhibits high precision (88.75%) and recall (69.44%) for forts features compared to recognize the house features. Therefore, the proposed strategy based on machine learning can be employed as an efficient method to assess architectural forms during the urban renewal process as well as a quantitative tool to extract aspects of regional architectures.
Architecture Heritage Recognition Using YOLACT Instance Segmentation
Bahrain has a rich architectural heritage that is reflected in its forts and houses. The forts and houses of Bahrain are not only important cultural landmarks but also serve as examples of the region's architectural styles. These forts and houses are built using locally-sourced materials, such as coral stone and palm wood, and feature distinctive design elements, such as wind towers, courtyards, and arched windows. Even, they serve as important cultural landmarks and provide valuable insights into the region's building techniques. However, the previous methods that are currently employed for architecture form come within the category of subjective analysis. In order to identify architecture form, identify regional architectural style, and provide a quantitative evaluation, quantitative approaches are essential. In this research study, a novel quantitative features based YOLACT object detection model has been employed. Firstly, the Bahrain architectural heritage forts and houses images have been collected from secondary sources with the ramparts, bastions, gateways, watch towers, large windows, high ceilings, wide front porches feature. Secondly, an image resizing and normalization techniques were used to preprocess the images. The preprocessed images have been used to train the object detection model that indentify the eight forts and houses features. Moreover, the YOLACT model obtained the 599 features of forts and 390 features of houses from preprocessed images directly with several epochs and learning rate. Furthermore, the performance of YOLACT model has been analyzed through precision, recall, F1-score, AP parameters. YOLACT model exhibits high precision (88.75%) and recall (69.44%) for forts features compared to recognize the house features. Therefore, the proposed strategy based on machine learning can be employed as an efficient method to assess architectural forms during the urban renewal process as well as a quantitative tool to extract aspects of regional architectures.
Architecture Heritage Recognition Using YOLACT Instance Segmentation
Kumar, Deepak (Autor:in) / Kukreja, Vinay (Autor:in) / Jain, Anuj kumar (Autor:in) / Bansal, Ankit (Autor:in)
03.08.2023
511030 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Instance segmentation using semi-supervised learning for fire recognition
| Elsevier | 2022