Performance/power assessment of CNN packages on embedded automotive platforms

TL;DR

This paper evaluates the performance and power efficiency of recent CNN models on embedded automotive platforms to guide engineers in selecting suitable systems for autonomous driving applications.

Contribution

It provides a comprehensive assessment of CNN packages on various state-of-the-art embedded hardware for automotive use, offering practical guidelines for system sizing and selection.

Findings

Recent CNN models achieve high accuracy and FPS on embedded platforms.

Power and size constraints significantly impact CNN deployment in automotive systems.

Guidelines are proposed for choosing optimal CNN packages and hardware configurations.

Abstract

The rise of power-efficient embedded computers based on highly-parallel accelerators opens a number of opportunities and challenges for researchers and engineers, and paved the way to the era of edge computing. At the same time, advances in embedded AI for object detection and categorization such as YOLO, GoogleNet and AlexNet reached an unprecedented level of accuracy (mean-Average Precision - mAP) and performance (Frames-Per-Second - FPS). Today, edge computers based on heterogeneous many-core systems are a predominant choice to deploy such systems in industry 4.0, wearable devices, and - our focus - autonomous driving systems. In these latter systems, engineers struggle to make reduced automotive power and size budgets co-exist with the accuracy and performance targets requested by autonomous driving. We aim at validating the effectiveness and efficiency of most recent networks on state-of-the-art platforms with embedded commercial-off-the-shelf System-on-Chips, such as Xavier AGX, Tegra X2 and Nano for NVIDIA and XCZU9EG and XCZU3EG of the Zynq UltraScale+ family, for the Xilinx counterpart. Our work aims at supporting engineers in choosing the most appropriate CNN package and computing system for their designs, and deriving guidelines for adequately sizing their systems.