Deep Learning Models Benchmarking for the Image Recognition

Authors

  • Bohdan Blagitko Dept. of Radiophysics and Computer Technologies, Ivan Franko National University of Lviv
  • Dmytro Myronyuk Dept. of Radiophysics and Computer Technologies, Ivan Franko National University of Lviv
  • Igor Zajachuk Pidstryhach Institute for Applied Problems of Mechanics and Mathematics National Academy of Sciences of Ukraine
  • Yurii Bodakovskyi Ukrainian Catholic University Faculty of Applied Sciences

Keywords:

deep learning, image recognition, computer vision, Raspberry Pi

Abstract

Machine learning has gained popularity for image recognition on edge devices. This study evaluates the performance of three devices, Nvidia Jetson Nano, Raspberry Pi 4 Model B, and Raspberry Pi 5 at accuracy, latency, and other metrics. The proprietary neural image recognition network is installed on each device, and their activity is analyzed. The study identifies the pros and cons of different image recognition methods. The various image recognition methods' advantages and disadvantages are identified.

Downloads

Published

2024-05-24

Issue

2024: INFORMATION TECHNOLOGIES AND COMPUTER MODELLING proceedings of the International Scientific Conference

Section

Section 4 Deep analysis and data organization, big data technologies, artificial intelligence systems, smart applications