eduzhai > Applied Sciences > Engineering >

Optimisation of the PointPillars network for 3D object detection in point clouds

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 7 pages

Abstract: In this paper we present our research on the optimisation of a deep neuralnetwork for 3D object detection in a point cloud. Techniques like quantisationand pruning available in the Brevitas and PyTorch tools were used. We performedthe experiments for the PointPillars network, which offers a reasonablecompromise between detection accuracy and calculation complexity. The aim ofthis work was to propose a variant of the network which we will ultimatelyimplement in an FPGA device. This will allow for real-time LiDAR dataprocessing with low energy consumption. The obtained results indicate that evena significant quantisation from 32-bit floating point to 2-bit integer in themain part of the algorithm, results in 5 -9 decrease of the detectionaccuracy, while allowing for almost a 16-fold reduction in size of the model.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...