ImageOP: The Image Dataset with Religious Buildings in the World Heritage Town of Ouro Preto for Deep Learning Classification DOI Creative Commons
André Luiz Carvalho Ottoni, Lara Toledo Cordeiro Ottoni

Heritage, Journal Year: 2024, Volume and Issue: 7(11), P. 6499 - 6525

Published: Nov. 20, 2024

Artificial intelligence has significant applications in computer vision studies for cultural heritage. In this research field, visual inspection of historical buildings and the digitization heritage using machine learning models stand out. However, literature still lacks datasets classification identification Brazilian religious deep learning, particularly with images from historic town Ouro Preto. It is noteworthy that Preto was first World Heritage Site recognized by UNESCO 1980. context, paper aims to address gap proposing a new image dataset, termed ImageOP: The Image Dataset Religious Buildings Town Deep Learning Classification. This dataset comprises 1613 facades 32 monuments Preto, categorized into five classes: fronton (pediment), door, window, tower, church. experiments validate ImageOP were conducted two stages: simulations smartphones. Furthermore, structures (MobileNet V2 EfficientNet B0) evaluated Edge Impulse software. MobileNet B0 are architectures convolutional neural networks designed aiming at low computational cost, real-time on mobile devices. results indicated utilizing achieved best outcomes simulations, accuracy = 94.5%, precision 96.0%, recall F-score 96.0%. Additionally, superior values obtained detecting (96.4%), church (97.1%), window (89.2%), door (94.7%), tower (95.4%). smartphones reinforced effectiveness proposed showing an average 88.0% building elements across nine tested device application. available Mendeley Data repository.

Language: Английский

ImageOP: The Image Dataset with Religious Buildings in the World Heritage Town of Ouro Preto for Deep Learning Classification DOI Creative Commons
André Luiz Carvalho Ottoni, Lara Toledo Cordeiro Ottoni

Heritage, Journal Year: 2024, Volume and Issue: 7(11), P. 6499 - 6525

Published: Nov. 20, 2024

Artificial intelligence has significant applications in computer vision studies for cultural heritage. In this research field, visual inspection of historical buildings and the digitization heritage using machine learning models stand out. However, literature still lacks datasets classification identification Brazilian religious deep learning, particularly with images from historic town Ouro Preto. It is noteworthy that Preto was first World Heritage Site recognized by UNESCO 1980. context, paper aims to address gap proposing a new image dataset, termed ImageOP: The Image Dataset Religious Buildings Town Deep Learning Classification. This dataset comprises 1613 facades 32 monuments Preto, categorized into five classes: fronton (pediment), door, window, tower, church. experiments validate ImageOP were conducted two stages: simulations smartphones. Furthermore, structures (MobileNet V2 EfficientNet B0) evaluated Edge Impulse software. MobileNet B0 are architectures convolutional neural networks designed aiming at low computational cost, real-time on mobile devices. results indicated utilizing achieved best outcomes simulations, accuracy = 94.5%, precision 96.0%, recall F-score 96.0%. Additionally, superior values obtained detecting (96.4%), church (97.1%), window (89.2%), door (94.7%), tower (95.4%). smartphones reinforced effectiveness proposed showing an average 88.0% building elements across nine tested device application. available Mendeley Data repository.

Language: Английский

Citations

1