Symmetry detection of occluded point cloud using deep learning
Zhelun Wu, Hongyan Jiang, Siyun He · Procedia Computer Science · 2021
Before our paper, some papers have approached symmetry detection in various attacks of lines. Deep learning has taken off in recent years in many fields in computer graphics, and we are thinking of using big data to solve the symmetry detection problem. Our work aims to solve a niche problem: using deep learning to detect symmetry in objects of the occluded point cloud. As far as we know, we are the first piece of work to deal with such a problem in deep learning settings. We employ points on the symmetry plane and normal vectors as double supervision to help us pinpoint the symmetry plane. Experiments conducted on the YCB-video dataset prove the effectiveness of the work and our method. To see our implementation, please visit https://github.com/Allen--Wu/dense_symmetry.