Evaluation of YOLOv7 Deployment Performance on Various Platforms (CPUs, GPUs) and Strategies

Authors

  • Fatima Abouzid University Sidi Mohamed Ben Abdellah https://orcid.org/0000-0002-3065-0923
  • Shadi Saleh Chemnitz University of Technology
  • Salim Harras Harras Chemnitz University of Technology
  • Wolfram Hardt Chemnitz University of Technology
  • Ghennioui Hicham University Sidi Mohamed Ben Abdellah

DOI:

https://doi.org/10.14464/ess.v10i8.660

Abstract

In the past few years, machine learning (ML) has evolved from academic research curiosity into a practical domain capable of addressing tangible business challenges.
Deploying machine learning (ML) models to production with an equivalent degree of thoroughness and automation as conventional software systems has proven to be a complex endeavor, demanding additional attention and infrastructure to address the unique challenges involved.
Building and training machine learning models is just one part of the process. Deploying models into production is a critical step that brings the potential of these models to life and allows them to deliver value to organizations and end users.
It's the bridge between the theoretical work in the research and the practical application of machine learning in real-world scenarios. This research paper presents an original study focused on deploying a traffic sign detection and recognition system utilizing the YOLOv7 algorithm. The study evaluates the algorithm's performance across diverse platforms and determines a robust production deployment strategy for YOLOv7.
Index Terms—Traffic sign, Detection, Classification, Deep Neural networks, YOLOv7, Deployment

ESS

Downloads

Published

2023-12-22