Abstract: Three-dimensional (3D) point clouds are essential for representing traffic ... which leverages an improved generative adversarial network to predict and correct missing portions of point ...

However, there are few methods to jointly represent 3D shapes from both point cloud and multi-view modal data. Therefore, we propose a guidance enhancement network (GENet) for 3D shape recognition ...