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

TL;DR

This paper explores optimizing the PointPillars 3D object detection network using quantisation and pruning techniques to enable real-time, low-energy LiDAR data processing on FPGA devices, with minimal accuracy loss.

Contribution

It introduces a quantised and pruned variant of PointPillars suitable for FPGA implementation, balancing detection accuracy and computational efficiency.

Findings

Quantisation from 32-bit to 2-bit reduces model size by 16 times.

Detection accuracy decreases by 5-9% with aggressive quantisation.

Optimized network enables real-time processing on FPGA with low energy consumption.

Abstract

In this paper we present our research on the optimisation of a deep neural network for 3D object detection in a point cloud. Techniques like quantisation and pruning available in the Brevitas and PyTorch tools were used. We performed the experiments for the PointPillars network, which offers a reasonable compromise between detection accuracy and calculation complexity. The aim of this work was to propose a variant of the network which we will ultimately implement in an FPGA device. This will allow for real-time LiDAR data processing with low energy consumption. The obtained results indicate that even a significant quantisation from 32-bit floating point to 2-bit integer in the main part of the algorithm, results in 5%-9% decrease of the detection accuracy, while allowing for almost a 16-fold reduction in size of the model.