Simultaneous Exploration of Geometric Information in 3d Model Construction Based on Parametric Models Using Point Cloud Data

Key Points

• The aim is to develop a method for efficiently constructing 3D models from point cloud data using parametric models.
• Utilized pre-trained neural networks based on parametric models.
• Simultaneously explored shape classes and shape parameters during model training.
• Conducted empirical analysis using simplified shape point cloud data of bridges.
• Confirmed that simultaneous exploration of shape classes and parameters influences model selection behavior.
• Demonstrated the parameter variability of the parametric models in practice.