| Title |
Leveraging Deep Learning and Path Planning Techniques for 3D ModelingFrom House Floor Plans |
| Authors |
미아오쉬(Miao, Xu) ; 엄신조(Eom, Shin-Jo) |
| DOI |
https://doi.org/10.5659/JAIK.2024.40.1.297 |
| Keywords |
Deep Learning; Path Planning Algorithm; Vectorization; Floor Plan Analysis; 3D modeling |
| Abstract |
The surge in BIM technology's popularity has heightened the demand for transforming traditional architectural floor plans into BIM-oriented
3D models. To address this, scholars have introduced diverse methods for this conversion, with those leveraging deep learning technology
garnering significant attention. Deep learning-based approaches typically encompass information segmentation and data vectorization. However,
the current research landscape lacks exploration into optimizing deep learning technology for high generalization efficiency and promptly
vectorizing wall lines. This study enhances the generalization capability of information segmentation in deep learning by incorporating Instance
Normalization (IN) and Instance Whitening (IW). Additionally, it introduces a method for swiftly generating wall lines through a proposed
vector generation algorithm rooted in path planning. This enhancement facilitates the creation of a BIM-centric automation process,
streamlining the conversion of accumulated 2D architectural drawings into efficient 3D models. |