Cover Image

Comparison of Deep Neural Networks Performance for Object Classification with Edge TPU

FYP 2 SEM 2 2019/2020

Saved in:
Bibliographic Details
Main Author: Ahmad Ammar Asyraaf Jainuddin
Format: text
Language:English
Published: 2023
Subjects:
Edge Computing
Deep Neural Network
Benchmarking
Tags: Add Tag
No Tags, Be the first to tag this record!
id my.uniten.dspace-21280
record_format dspace
spelling my.uniten.dspace-212802023-12-06T14:29:29Z Comparison of Deep Neural Networks Performance for Object Classification with Edge TPU Ahmad Ammar Asyraaf Jainuddin Edge Computing Deep Neural Network Benchmarking FYP 2 SEM 2 2019/2020 As machine learning (ML) becomes more widespread, the speed of data send to the processing unit can be crucial depending on application. The processing unit of a ML applications are usually centralize and this cause speed of data transfer to increase as data gathering devices were installed further away from the processing unit. This thesis aims to provide the public with information and data on Google’s new machine learning hardware called Edge TPU that was created specifically for edge devices and of the author’s analysis and review of the data gathered from this research. For this purpose, many deep neural network (DNN) models were sought and reviewed. Then parts of this pool of DNN models were chosen to be benchmark with various hardware used in edge application. First, due to the seeking of suitable models and edge hardware for benchmarking, an information available online was used to see the popularity of each neural network models and check with Tensorflow’s own benchmarking results of various DNN model and edge devices in order to be able to compare results. Programming languages available are also studied in the same way to ease the programming of benchmarking codes and the results of the benchmark are pass through a mathematical formula for better understanding of data. This studies find that the Edge TPU allows edge devices to run DNN models at a faster speed than without using the Edge TPU. The studies also shows which DNN models are more suitable for edge devices. 2023-05-03T16:26:34Z 2023-05-03T16:26:34Z 2020-02 Resource Types::text https://irepository.uniten.edu.my/handle/123456789/21280 en application/pdf
institution Universiti Tenaga Nasional
building UNITEN Library
collection Institutional Repository
continent Asia
country Malaysia
content_provider Universiti Tenaga Nasional
content_source UNITEN Institutional Repository
url_provider http://dspace.uniten.edu.my/
language English
topic Edge Computing
Deep Neural Network
Benchmarking
spellingShingle Edge Computing
Deep Neural Network
Benchmarking
Ahmad Ammar Asyraaf Jainuddin
Comparison of Deep Neural Networks Performance for Object Classification with Edge TPU
description FYP 2 SEM 2 2019/2020
format Resource Types::text
author Ahmad Ammar Asyraaf Jainuddin
author_facet Ahmad Ammar Asyraaf Jainuddin
author_sort Ahmad Ammar Asyraaf Jainuddin
title Comparison of Deep Neural Networks Performance for Object Classification with Edge TPU
title_short Comparison of Deep Neural Networks Performance for Object Classification with Edge TPU
title_full Comparison of Deep Neural Networks Performance for Object Classification with Edge TPU
title_fullStr Comparison of Deep Neural Networks Performance for Object Classification with Edge TPU
title_full_unstemmed Comparison of Deep Neural Networks Performance for Object Classification with Edge TPU
title_sort comparison of deep neural networks performance for object classification with edge tpu
publishDate 2023
_version_ 1806427762769002496
score 13.214268

Similar Items