19 March 2024 | Yiming Qin, Nanxuan Zhao, Jiale Yang, Siyuan Pan, Bin Sheng, Rynson W. H. Lau
The paper "UrbanEvolver: Function-Aware Urban Layout Regeneration" addresses the challenge of urban regeneration, specifically focusing on the task of function-aware urban layout regeneration. The authors propose UrbanEvolver, a deep generative model that aims to regenerate urban layouts (roads and buildings) for a target region based on its function and surrounding context. The model extracts implicit regeneration rules from the target function and surrounding context using function-layout adaptive (FA) blocks, which encode these rules at different scales. To ensure the validity and adherence to road structures, UrbanEvolver incorporates a set of losses that cover both pixel-level and geometry-level constraints. The model is trained on a large-scale urban layout dataset containing over 147,000 regions with rich annotations. Extensive experiments demonstrate that UrbanEvolver outperforms baseline methods in generating practical and function-aware urban layouts. Key contributions include the introduction of UrbanEvolver and the FA block, as well as the development of a comprehensive set of loss functions.The paper "UrbanEvolver: Function-Aware Urban Layout Regeneration" addresses the challenge of urban regeneration, specifically focusing on the task of function-aware urban layout regeneration. The authors propose UrbanEvolver, a deep generative model that aims to regenerate urban layouts (roads and buildings) for a target region based on its function and surrounding context. The model extracts implicit regeneration rules from the target function and surrounding context using function-layout adaptive (FA) blocks, which encode these rules at different scales. To ensure the validity and adherence to road structures, UrbanEvolver incorporates a set of losses that cover both pixel-level and geometry-level constraints. The model is trained on a large-scale urban layout dataset containing over 147,000 regions with rich annotations. Extensive experiments demonstrate that UrbanEvolver outperforms baseline methods in generating practical and function-aware urban layouts. Key contributions include the introduction of UrbanEvolver and the FA block, as well as the development of a comprehensive set of loss functions.