Основні поняття
Deep learning techniques, particularly deep generative models, enable the efficient and diverse generation of virtual architecture to enrich the content and user experiences in the metaverse.
Анотація
This survey provides a comprehensive overview of the current state of research on leveraging deep learning for generating virtual architecture. It covers the following key aspects:
-
Virtual Architecture and Design Discipline:
- Defines virtual architecture as a spatial instance within the virtual world, characterized by interactive features related to social attributions and technology frameworks.
- Highlights the overlapping scope between virtual architecture and human-building interaction (HBI), emphasizing the importance of considering user presence, socialization, interactivity, and interoperability.
- Outlines the design considerations for virtual architecture, including building form, production mode, and the pivotal role of deep learning approaches.
-
Deep Generative Models and 3D Representations:
- Introduces the progression of deep generative models, such as GANs, VAEs, and diffusion models, for 3D shape generation and 3D-aware image synthesis.
- Discusses various 3D representations, including voxel grids, point clouds, meshes, and neural fields, and their applications in architectural design.
-
Generated Virtual Architecture by Deep Learning:
- Reviews the current approaches to generating virtual architecture using deep learning, including 3D transposition and 3D solid form generation.
- Identifies four key focuses in the literature: dataset, multimodality, design intuition, and generative framework.
-
Generation Approaches for Virtual Architecture:
- Investigates the generative approaches of virtual architecture utilizing various deep generative models, including GAN, VAE, 3D-aware image synthesis, and diffusion models.
- Discusses the characteristics and capabilities of these approaches, as well as their challenges and potential applications in virtual architecture design.
The survey highlights the importance of considering human and social factors in the development of automatically generated virtual architecture, emphasizing the need for innovative methods that prioritize user experience and collaboration between designers and deep learning techniques.