Vision‐based adaptive cross‐domain online product recommendation for 3D design models DOI

Xiaoping Zhou,

Qin Si,

Gen Liu

et al.

Computer-Aided Civil and Infrastructure Engineering, Journal Year: 2025, Volume and Issue: unknown

Published: May 13, 2025

Abstract Three‐dimensional (3D) digital design is extensively adopted in the architecture, engineering, consulting, operations, and maintenance (AECOM) industry to enhance collaboration among stakeholders. Although recommendation systems are commonly employed facilitate purchasing e‐commerce websites, none involves recommending online products users from 3D building models due dimensional stylistic discrepancies. This study proposes a vision‐based adaptive cross‐domain product method, VacRed, for models. First, approach proposed transform into images, addressing discrepancies dimension style between them. Second, an mechanism introduced solve issue of image quality instability caused by variations generator weights during training process generative Third, system developed based on deep learning recommend top k relevant given product. Finally, experiments were conducted ascertain effectiveness VacRed method. The experimental results this method demonstrate its excellent performance, achieving precision rate ( PR ) 87.20% mean average 83.65%. effectively connects two main stages AECOM industry, purchasing, large communities, e‐commerce.

Language: Английский

Vision‐based adaptive cross‐domain online product recommendation for 3D design models DOI

Xiaoping Zhou,

Qin Si,

Gen Liu

et al.

Computer-Aided Civil and Infrastructure Engineering, Journal Year: 2025, Volume and Issue: unknown

Published: May 13, 2025

Abstract Three‐dimensional (3D) digital design is extensively adopted in the architecture, engineering, consulting, operations, and maintenance (AECOM) industry to enhance collaboration among stakeholders. Although recommendation systems are commonly employed facilitate purchasing e‐commerce websites, none involves recommending online products users from 3D building models due dimensional stylistic discrepancies. This study proposes a vision‐based adaptive cross‐domain product method, VacRed, for models. First, approach proposed transform into images, addressing discrepancies dimension style between them. Second, an mechanism introduced solve issue of image quality instability caused by variations generator weights during training process generative Third, system developed based on deep learning recommend top k relevant given product. Finally, experiments were conducted ascertain effectiveness VacRed method. The experimental results this method demonstrate its excellent performance, achieving precision rate ( PR ) 87.20% mean average 83.65%. effectively connects two main stages AECOM industry, purchasing, large communities, e‐commerce.

Language: Английский

Citations

0