A deep neural network for detecting obstacle instances using radar sensors in autonomous machine applications

摘要

Problem to be solved: to provide object detection for an autonomous machine using a deep neural network (DNN).DNN is trained to detect moving and still obstacles from 3D spatial radar data.The ground truth tracing data of DNN is generated from lidar data, and the scene is observed by radar and lidar sensors for the purpose of collecting radar and lidar data of a specific time slice.Radar data is used for input training data, and lidar data associated with the same or nearest time slice as radar data is annotated with a ground truth label identifying the object to be detected.The lidar label is transmitted to radar data, and the lidar label containing radar detection less than the threshold value is removed and the remaining lidar labels are used for ground truth data generation.Diagram