Hybrid Optimization of Target Detection on Embedded Platforms for Real Time Applications

Xinchen Zhang, Wangchao Sun, Yaodong Zhao, Kaisheng Liao, Yilin Liu, Hongda Xu, Zhuoling Xiao*, Bo Yan

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Target detection has been widely used in fields such as intelligent security and autonomous driving. However, existing computationally heavy target detection algorithms based on deep learning can only work on GPU and CPU platforms, restricting the applications on edge devices with limited computational power. To address this issue, this paper proposes layer fusion and 16-bit fixed-point quantization on the YOLOv2-Tiny algorithm to reduce the computational complexity of target detection algorithms. Furthermore, the data transmission efficiency is optimized by using ping-pong butter and multi-channel methods. To reduce FPGA resource consumption, the neural network is split into convolution, accumulation, pooling, and address mapping modules. The proposed system has been successfully implemented on the Xilinx Zynq-XC7Z035 platform, using only 47% of BRAM resources and 18% of DSP resources.

Original languageEnglish
Title of host publication2022 IEEE 5th International Conference on Electronics Technology, ICET 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages1136-1141
Number of pages6
ISBN (Electronic)9781665485081
DOIs
Publication statusPublished - 2022
Externally publishedYes
Event5th IEEE International Conference on Electronics Technology, ICET 2022 - Chengdu, China
Duration: 13 May 202216 May 2022

Publication series

Name2022 IEEE 5th International Conference on Electronics Technology, ICET 2022

Conference

Conference5th IEEE International Conference on Electronics Technology, ICET 2022
Country/TerritoryChina
CityChengdu
Period13/05/2216/05/22

Keywords

  • target detection
  • YOLOv2-Tiny
  • Zynq

Fingerprint

Dive into the research topics of 'Hybrid Optimization of Target Detection on Embedded Platforms for Real Time Applications'. Together they form a unique fingerprint.

Cite this