Parallel Point Clouds: Hybrid Point Cloud Generation and 3D Model Enhancement via Virtual–Real Integration

Three-dimensional information perception from point clouds is of vital importance for improving the ability machines to understand world, especially autonomous driving and unmanned aerial vehicles. Data annotation one most challenging costly tasks. In this paper, we propose a closed-loop virtual–real interactive cloud generation model-upgrading framework called Parallel Point Clouds (PPCs). To our best knowledge, first time that training model has been changed an open-loop mechanism. The feedback evaluation results used update dataset, benefiting flexibility artificial scenes. Under framework, point-based LiDAR simulation proposed, which greatly simplifies scanning operation. Besides, group-based placing method put forward integrate hybrid clouds, via locating candidate positions virtual objects in real Taking advantage CAD models mobile devices, two datasets, i.e., ShapeKITTI MobilePointClouds, are built 3D detection With almost zero labor cost on data newly added objects, (PointPillars) trained with MobilePointClouds achieved 78.6% 60.0% average precision detection, respectively.