Optimisation Techniques for Compact CNN on Embedded Systems for Gesture Recognition

Embedded applications are increasingly prevalent in various domains, from consumer electronics to industrial automation and smart cities. With the advances in integrated circuit manufacturing technologies, low-power chips can now execute complex algorithms, including machine learning models. Howeve...

Full description

Saved in:
Bibliographic Details
Main Authors: João Carlos Bittencourt, Walber Conceição de Jesus Rocha
Format: Article
Language:English
Published: Universidade do Porto 2023-11-01
Series:U.Porto Journal of Engineering
Subjects:
Online Access:https://journalengineering.fe.up.pt/index.php/upjeng/article/view/2156
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849320013699219456
author João Carlos Bittencourt
Walber Conceição de Jesus Rocha
author_facet João Carlos Bittencourt
Walber Conceição de Jesus Rocha
author_sort João Carlos Bittencourt
collection DOAJ
description Embedded applications are increasingly prevalent in various domains, from consumer electronics to industrial automation and smart cities. With the advances in integrated circuit manufacturing technologies, low-power chips can now execute complex algorithms, including machine learning models. However, the computational constraints of embedded devices require compact and efficient neural network models, as well as software frameworks and optimisation techniques tailored to their hardware resources. This study investigates the implementation of Convolutional Neural Network (CNN) models for gesture recognition on an STM32F4 microcontroller, by exploring the impact of freezing layers, fine-tuning and pruning techniques on pre-trained CNN models. The results demonstrate that fine-tuning and freezing layers improve accuracy by up to 18%. Additionally, freezing layers by 10% and 20% improved the accuracy. Finally, we demonstrate that pruning reduced the model size by 90%, enabling it to perform gesture recognition on small devices. These findings are significant for developing software and optimisation techniques for embedded systems, particularly in the context of the Internet of Things.
format Article
id doaj-art-3a80331ced05451b8f57b706fca31f77
institution Kabale University
issn 2183-6493
language English
publishDate 2023-11-01
publisher Universidade do Porto
record_format Article
series U.Porto Journal of Engineering
spelling doaj-art-3a80331ced05451b8f57b706fca31f772025-08-20T03:50:16ZengUniversidade do PortoU.Porto Journal of Engineering2183-64932023-11-019510.24840/2183-6493_009-005_002156Optimisation Techniques for Compact CNN on Embedded Systems for Gesture RecognitionJoão Carlos Bittencourt0https://orcid.org/0000-0002-4540-512XWalber Conceição de Jesus Rocha1University of Porto, Faculty of EngineeringFederal University of Recôncavo da Bahia, Centre for Exact and Technological Sciences Embedded applications are increasingly prevalent in various domains, from consumer electronics to industrial automation and smart cities. With the advances in integrated circuit manufacturing technologies, low-power chips can now execute complex algorithms, including machine learning models. However, the computational constraints of embedded devices require compact and efficient neural network models, as well as software frameworks and optimisation techniques tailored to their hardware resources. This study investigates the implementation of Convolutional Neural Network (CNN) models for gesture recognition on an STM32F4 microcontroller, by exploring the impact of freezing layers, fine-tuning and pruning techniques on pre-trained CNN models. The results demonstrate that fine-tuning and freezing layers improve accuracy by up to 18%. Additionally, freezing layers by 10% and 20% improved the accuracy. Finally, we demonstrate that pruning reduced the model size by 90%, enabling it to perform gesture recognition on small devices. These findings are significant for developing software and optimisation techniques for embedded systems, particularly in the context of the Internet of Things. https://journalengineering.fe.up.pt/index.php/upjeng/article/view/2156Embedded systemsTinyMLInternet of ThingsConvolutional Neural NetworksMobileNetEfficientNet
spellingShingle João Carlos Bittencourt
Walber Conceição de Jesus Rocha
Optimisation Techniques for Compact CNN on Embedded Systems for Gesture Recognition
U.Porto Journal of Engineering
Embedded systems
TinyML
Internet of Things
Convolutional Neural Networks
MobileNet
EfficientNet
title Optimisation Techniques for Compact CNN on Embedded Systems for Gesture Recognition
title_full Optimisation Techniques for Compact CNN on Embedded Systems for Gesture Recognition
title_fullStr Optimisation Techniques for Compact CNN on Embedded Systems for Gesture Recognition
title_full_unstemmed Optimisation Techniques for Compact CNN on Embedded Systems for Gesture Recognition
title_short Optimisation Techniques for Compact CNN on Embedded Systems for Gesture Recognition
title_sort optimisation techniques for compact cnn on embedded systems for gesture recognition
topic Embedded systems
TinyML
Internet of Things
Convolutional Neural Networks
MobileNet
EfficientNet
url https://journalengineering.fe.up.pt/index.php/upjeng/article/view/2156
work_keys_str_mv AT joaocarlosbittencourt optimisationtechniquesforcompactcnnonembeddedsystemsforgesturerecognition
AT walberconceicaodejesusrocha optimisationtechniquesforcompactcnnonembeddedsystemsforgesturerecognition