Edge-Optimized Deep Learning Architectures for Classification of Agricultural Insects with Mobile Deployment

The deployment of machine learning models on mobile platforms has ushered in a new era of innovation across diverse sectors, including agriculture, where such applications hold immense promise for empowering farmers with cutting-edge technologies. In this context, the threat posed by insects to crop...

Full description

Saved in:
Bibliographic Details
Main Author: Akhtar, Muhammad Hannan (author)
Other Authors: Eksheir, Ibrahim (author), Shanableh, Tamer (author)
Format: article
Published: 2025
Subjects:
Online Access:https://hdl.handle.net/11073/26040
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1864513435819048960
author Akhtar, Muhammad Hannan
author2 Eksheir, Ibrahim
Shanableh, Tamer
author2_role author
author
author_facet Akhtar, Muhammad Hannan
Eksheir, Ibrahim
Shanableh, Tamer
author_role author
dc.creator.none.fl_str_mv Akhtar, Muhammad Hannan
Eksheir, Ibrahim
Shanableh, Tamer
dc.date.none.fl_str_mv 2025-05-01T08:44:13Z
2025-05-01T08:44:13Z
2025-04-25
dc.format.none.fl_str_mv application/pdf
dc.identifier.none.fl_str_mv Akhtar,M.H.;Eksheir,I.; Shanableh, T. Edge-OptimizedDeep Learning Architectures for Classification of Agricultural Insects withMobileDeployment. Information 2025, 16, 348. https://doi.org/10.3390/info16050348
2078-2489
https://hdl.handle.net/11073/26040
10.3390/info16050348
dc.language.none.fl_str_mv en_US
dc.publisher.none.fl_str_mv MDPI
dc.relation.none.fl_str_mv https://doi.org/10.3390/info16050348
dc.rights.none.fl_str_mv Attribution 4.0 International
http://creativecommons.org/licenses/by/4.0/
dc.subject.none.fl_str_mv On-edge classification
Model quantization
TensorFlow Lite
Insect classification
EfficientNet
Deep learning
dc.title.none.fl_str_mv Edge-Optimized Deep Learning Architectures for Classification of Agricultural Insects with Mobile Deployment
dc.type.none.fl_str_mv Published version
Peer-Reviewed
info:eu-repo/semantics/publishedVersion
info:eu-repo/semantics/article
description The deployment of machine learning models on mobile platforms has ushered in a new era of innovation across diverse sectors, including agriculture, where such applications hold immense promise for empowering farmers with cutting-edge technologies. In this context, the threat posed by insects to crop yields during harvest has escalated, fueled by factors such as evolution and climate change-induced shifts in insect behavior. To address this challenge, smart insect monitoring systems and detection models have emerged as crucial tools for farmers and IoT-based systems, enabling interventions to safeguard crops. The primary contribution of this study lies in its systematic investigation of model optimization techniques for edge deployment, including Post-Training Quantization, Quantization-Aware Training, and Data Representative Quantization. As such, we address the crucial need for efficient, on-site pest detection tools in agricultural settings. We provide a detailed analysis of the trade-offs between model size, inference speed, and accuracy across different optimization approaches, ensuring practical applicability in resource-constrained farming environments. Our study explores various methodologies for model development, including the utilization of Mobile-ViT and EfficientNet architectures, coupled with transfer learning and fine-tuning techniques. Using the Dangerous Farm Insects Dataset, we achieve an accuracy of 82.6% and 77.8% on validation and test datasets, respectively, showcasing the efficacy of our approach. Furthermore, we investigate quantization techniques to optimize model performance for on-device inference, ensuring seamless deployment on mobile devices and other edge devices without compromising accuracy. The best quantized model, produced through Post-Training Quantization, was able to maintain a classification accuracy of 77.8% while significantly reducing the model size from 33 MB to 9.6 MB. To validate the generalizability of our solution, we extended our experiments to the larger IP102 dataset. The quantized model produced using Post-Training Quantization was able to maintain a classification accuracy of 59.6% while also reducing the model size from 33 MB to 9.6 MB, thus demonstrating that our solution maintains a competitive performance across a broader range of insect classes.
format article
id aus_0d8e4cb4cd3890f1c19cec906e50b693
identifier_str_mv Akhtar,M.H.;Eksheir,I.; Shanableh, T. Edge-OptimizedDeep Learning Architectures for Classification of Agricultural Insects withMobileDeployment. Information 2025, 16, 348. https://doi.org/10.3390/info16050348
2078-2489
10.3390/info16050348
language_invalid_str_mv en_US
network_acronym_str aus
network_name_str aus
oai_identifier_str oai:repository.aus.edu:11073/26040
publishDate 2025
publisher.none.fl_str_mv MDPI
repository.mail.fl_str_mv
repository.name.fl_str_mv
repository_id_str
rights_invalid_str_mv Attribution 4.0 International
http://creativecommons.org/licenses/by/4.0/
spelling Edge-Optimized Deep Learning Architectures for Classification of Agricultural Insects with Mobile DeploymentAkhtar, Muhammad HannanEksheir, IbrahimShanableh, TamerOn-edge classificationModel quantizationTensorFlow LiteInsect classificationEfficientNetDeep learningThe deployment of machine learning models on mobile platforms has ushered in a new era of innovation across diverse sectors, including agriculture, where such applications hold immense promise for empowering farmers with cutting-edge technologies. In this context, the threat posed by insects to crop yields during harvest has escalated, fueled by factors such as evolution and climate change-induced shifts in insect behavior. To address this challenge, smart insect monitoring systems and detection models have emerged as crucial tools for farmers and IoT-based systems, enabling interventions to safeguard crops. The primary contribution of this study lies in its systematic investigation of model optimization techniques for edge deployment, including Post-Training Quantization, Quantization-Aware Training, and Data Representative Quantization. As such, we address the crucial need for efficient, on-site pest detection tools in agricultural settings. We provide a detailed analysis of the trade-offs between model size, inference speed, and accuracy across different optimization approaches, ensuring practical applicability in resource-constrained farming environments. Our study explores various methodologies for model development, including the utilization of Mobile-ViT and EfficientNet architectures, coupled with transfer learning and fine-tuning techniques. Using the Dangerous Farm Insects Dataset, we achieve an accuracy of 82.6% and 77.8% on validation and test datasets, respectively, showcasing the efficacy of our approach. Furthermore, we investigate quantization techniques to optimize model performance for on-device inference, ensuring seamless deployment on mobile devices and other edge devices without compromising accuracy. The best quantized model, produced through Post-Training Quantization, was able to maintain a classification accuracy of 77.8% while significantly reducing the model size from 33 MB to 9.6 MB. To validate the generalizability of our solution, we extended our experiments to the larger IP102 dataset. The quantized model produced using Post-Training Quantization was able to maintain a classification accuracy of 59.6% while also reducing the model size from 33 MB to 9.6 MB, thus demonstrating that our solution maintains a competitive performance across a broader range of insect classes.College of EngineeringDepartment of Computer Science and EngineeringMDPI2025-05-01T08:44:13Z2025-05-01T08:44:13Z2025-04-25Published versionPeer-Reviewedinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/articleapplication/pdfAkhtar,M.H.;Eksheir,I.; Shanableh, T. Edge-OptimizedDeep Learning Architectures for Classification of Agricultural Insects withMobileDeployment. Information 2025, 16, 348. https://doi.org/10.3390/info160503482078-2489https://hdl.handle.net/11073/2604010.3390/info16050348en_UShttps://doi.org/10.3390/info16050348Attribution 4.0 Internationalhttp://creativecommons.org/licenses/by/4.0/oai:repository.aus.edu:11073/260402025-05-01T14:40:57Z
spellingShingle Edge-Optimized Deep Learning Architectures for Classification of Agricultural Insects with Mobile Deployment
Akhtar, Muhammad Hannan
On-edge classification
Model quantization
TensorFlow Lite
Insect classification
EfficientNet
Deep learning
status_str publishedVersion
title Edge-Optimized Deep Learning Architectures for Classification of Agricultural Insects with Mobile Deployment
title_full Edge-Optimized Deep Learning Architectures for Classification of Agricultural Insects with Mobile Deployment
title_fullStr Edge-Optimized Deep Learning Architectures for Classification of Agricultural Insects with Mobile Deployment
title_full_unstemmed Edge-Optimized Deep Learning Architectures for Classification of Agricultural Insects with Mobile Deployment
title_short Edge-Optimized Deep Learning Architectures for Classification of Agricultural Insects with Mobile Deployment
title_sort Edge-Optimized Deep Learning Architectures for Classification of Agricultural Insects with Mobile Deployment
topic On-edge classification
Model quantization
TensorFlow Lite
Insect classification
EfficientNet
Deep learning
url https://hdl.handle.net/11073/26040