Contrastive Learning with Image Deformation and Refined NT-Xent Loss for Urban Morphology Discovery DOI Creative Commons
Chia‐Ho Hua,

Daijun Chen,

Meng Niu

et al.

ISPRS International Journal of Geo-Information, Journal Year: 2025, Volume and Issue: 14(5), P. 196 - 196

Published: May 8, 2025

The traditional paradigm for studying urban morphology involves the interpretation of Nolli maps, using methods such as morphometrics and visual neural networks. Previous studies on discovery have always been based raster analysis limited to central city area. Raster can lead fragmented forms, focusing only area ignores many representative forms in suburbs towns. In this study, a vast complex dataset was applied administrative community or village boundary, new image deformation pipeline proposed enhance morphological characteristics building groups. This allows networks focus extracting Additionally, research often uses unsupervised learning, which means that learning process is difficult control. Therefore, we refined NT-Xent loss so it integrate indicators. improvement network “recognize” similarity samples during optimization. By defining similarity, guide bring closer move them farther apart certain Three Chinese cities were used our testing. Representative types identified, particularly some located at fringe. data demonstrated effectiveness function, sociological illustrated unique functions these types.

Language: Английский

An approach for heritage settlement classification using block patterns: The study of 37 typical colonial heritage settlements in the Americas DOI Creative Commons

Wei Wei,

Liyang Liu,

Zhaosong Niu

et al.

Frontiers of Architectural Research, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 1, 2025

Language: Английский

Citations

0
Contrastive Learning with Image Deformation and Refined NT-Xent Loss for Urban Morphology Discovery DOI Creative Commons
Chia‐Ho Hua,

Daijun Chen,

Meng Niu

et al.

ISPRS International Journal of Geo-Information, Journal Year: 2025, Volume and Issue: 14(5), P. 196 - 196

Published: May 8, 2025

The traditional paradigm for studying urban morphology involves the interpretation of Nolli maps, using methods such as morphometrics and visual neural networks. Previous studies on discovery have always been based raster analysis limited to central city area. Raster can lead fragmented forms, focusing only area ignores many representative forms in suburbs towns. In this study, a vast complex dataset was applied administrative community or village boundary, new image deformation pipeline proposed enhance morphological characteristics building groups. This allows networks focus extracting Additionally, research often uses unsupervised learning, which means that learning process is difficult control. Therefore, we refined NT-Xent loss so it integrate indicators. improvement network “recognize” similarity samples during optimization. By defining similarity, guide bring closer move them farther apart certain Three Chinese cities were used our testing. Representative types identified, particularly some located at fringe. data demonstrated effectiveness function, sociological illustrated unique functions these types.

Language: Английский

Citations

0