A Learning Method for Reconstructing 3D Models from Sketches

Masaji Tanaka, Motoki Terano, Chiharu Higashino, Tetsuya Asano, Keigo Takasugi

Machine Learning, Reconstruction, Sketch, 3D Model, Line Drawing
Presently solid modelers have become popular tools in CAD/CAM systems, and three-dimensional (3D) sketching systems as simpler modelers have been developed in recent years. However, various basic operations are necessary to create solid models, making them difficult to use by beginners. Generally, sketches are convenient for expressing 3D models and most people can draw sketches. We have been developing a system that can automatically reconstruct 3D models from sketches that include curved lines. In this system, many of broken sketches be restored, so a learning method was introduced, but we could not explain its algorithm in detail. In this paper, an epochal learning method is proposed where a larger number of broken sketches be handled than by our previous method. Moreover, rough sketches can be corrected into precise sketches, and several examples are presented in this paper to demonstrate this method.

Computer-Aided Design & Applications

2019 CAD Solutions, LLC, http://www.cad-journal.net

16

6

1158

1170