Timezone: »

 
Poster
SLIBO-Net: Floorplan Reconstruction via Slicing Box Representation with Local Geometry Regularization
Jheng-Wei Su · Kuei-Yu Tung · Chi-Han Peng · Peter Wonka · Hung-Kuo (James) Chu

Wed Dec 13 08:45 AM -- 10:45 AM (PST) @ Great Hall & Hall B1+B2 #211
in Poster Session 3 »

This paper focuses on improving the reconstruction of 2D floorplans from unstructured 3D point clouds. We identify opportunities for enhancement over the existing methods in three main areas: semantic quality, efficient representation, and local geometric details. To address these, we presents SLIBO-Net, an innovative approach to reconstructing 2D floorplans from unstructured 3D point clouds. We propose a novel transformer-based architecture that employs an efficient floorplan representation, providing improved room shape supervision and allowing for manageable token numbers. By incorporating geometric priors as a regularization mechanism and post-processing step, we enhance the capture of local geometric details. We also propose a scale-independent evaluation metric, correcting the discrepancy in error treatment between varying floorplan sizes. Our approach notably achieves a new state-of-the-art on the Structure3D dataset. The resultant floorplans exhibit enhanced semantic plausibility, substantially improving the overall quality and realism of the reconstructions. Our code and dataset are available online.

Author Information

Jheng-Wei Su (National Tsing Hua University)
Kuei-Yu Tung (Department of Computer Science, National Tsing Hua University, National Tsinghua University)
Chi-Han Peng (National Yang Ming Chiao Tung University)
Chi-Han Peng

I am an Assistant Professor at the College of AI, National Yang Ming Chiao Tung University (NYCU) Tainan campus. My research areas include AI/neural networks, Computer Graphics, and Computer Vision in general. In particular, I am interested in 3D applications enabled by 360-degree panoramic cameras, AI in medicine, and AI in the architecture, engineering, and construction (AEC) industries. I was a research scientist at the Visual Computing Center at King Abdullah University of Science and Technology (KAUST), working with Prof. Peter Wonka. I was a postdoctoral research associate at University College London (UCL), working with Prof. Niloy Mitra. I worked at Nogle, a private company based in Taipei, on the 3D modelling and simulation of real-world clothes for the fashion / garment industries. I received PhD in Computer Science at Arizona State University. My thesis, Connectivity Control for Quad-Dominant Meshes, was supervised by Prof. Peter Wonka. During my PhD studies, I also took summer intern at Adobe ATL (supervised by Dr. Nathan Carr and Radomir Mech) and visited KAUST, Saudi Arabia, and TU Vienna, Austria, for research collaborations. Prior to my PhD study, I received my B.S. and M.S. in Computer Science at National Chiao Tung University (NCTU), Taiwan. I worked at CyberLink, the maker of PowerDVD video players, as a software engineer for four years. I was also the co-founder of an online taxi carpooling service serving the Taipei metropolitan area.

Peter Wonka (KAUST)
Hung-Kuo (James) Chu (National Tsing Hua University)
Hung-Kuo (James) Chu

Hung-Kuo Chu is a professor at the Department of Computer Science, National Tsing Hua University. He received BS and Ph.D. degrees from the Department of Computer Science and Information Engineering, National Cheng-Kung University. His research interests focus on image-based modeling, neural rendering, deep image/video synthesis, texture synthesis and inpainting, digital-twins simulation, and virtual/augmented reality.

More from the Same Authors