Optimization of PointPillars (A Deep Learning Network for LiDAR-based 3D Object Detection) on Intel Platform
Shengxian Liu, Qing Xu, Hua Ma, Jessica Du, Ming Lei, Jian Tao, Jiulong Bao · 2021 IEEE International Conference on Power, Intelligent Computing and Systems (ICPICS) · 2021
This paper introduces how to optimize the PointPillars, a network for the deep-learning-based object detection in 3D point clouds, on the 11th-Generation Intel® CoreTMProcessors (Tiger Lake) by using the Intel® Distribution of OpenVINOTMToolkit. The throughput requirement for the use cases of transportation infrastructure (e.g., 3D point clouds generated by the roadside Lidars) is 10 frames per second. In comparison with the existing solutions that we are aware of, our solution can achieve the throughput of 11.1 FPS and the latency of 154.7 ms on Intel® CoreTMprocessors with much lower cost and much lower power consumption.