A Computational-Efficient Deformable Convolution Network Accelerator via Hardware and Algorithm Co-Optimization

Shan Li, Shan Cao*, Lanqing Hui, Zhiyuan Jiang, Yanzan Sun, Shugong Xu

*Corresponding author for this work

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

5 Citations (Scopus)

Abstract

Deformable convolution networks (DCNs) shows performance boosts on object recognition tasks by enabling variable geometric modeling. However, the irregular addressing of memory accesses makes it inefficient for hardware acceleration. In this paper, we propose a computational-efficient hardware accelerator for DCNs. First, a hardware-friendly DCNs inference scheme is introduced based on the original DCNs algorithm with little accuracy loss. Secondly, a hardware accelerator architecture is presented correspondingly, and an speed matching method is introduced to maximizing the number of deformable layers without latency increase. The proposed accelerator is implemented on the Arria 10 FPGA, results of which show that the proposed design achieves the highest throughput and DSP efficiency compared with state-of-the-art DCNs accelerators.

Original languageEnglish
Title of host publication2022 IEEE Workshop on Signal Processing Systems, SiPS 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781665485241
DOIs
Publication statusPublished - 2022
Externally publishedYes
Event36th IEEE Workshop on Signal Processing Systems, SiPS 2022 - Rennes, France
Duration: 2 Nov 20224 Nov 2022

Publication series

NameIEEE Workshop on Signal Processing Systems, SiPS: Design and Implementation
Volume2022-November
ISSN (Print)1520-6130

Conference

Conference36th IEEE Workshop on Signal Processing Systems, SiPS 2022
Country/TerritoryFrance
CityRennes
Period2/11/224/11/22

Keywords

  • deformable convolution
  • FPGA
  • Hardware accelerator
  • hardware-friendly algorithm

Fingerprint

Dive into the research topics of 'A Computational-Efficient Deformable Convolution Network Accelerator via Hardware and Algorithm Co-Optimization'. Together they form a unique fingerprint.

Cite this