This is a Demo Server. Data inside this system is only for test purpose.

Izmir Institute of Technology

Permanent URI for this communityhttps://premium.gcris.co/handle/123456789/1

Browse

Browsing Izmir Institute of Technology by Author "Avcı, İsa"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • Thumbnail Image
    Conference Object
    A Lightweight Mobile Deep Learning Framework for Real-Time Plant Disease Detection in Smart Agriculture
    (Institute of Electrical and Electronics Engineers Inc., 2025) Avcı, İsa; Koca, Murat; Khan, Yahya Zakrya
    It is imperative to detect plant diseases early to enhance agricultural productivity and ensure food security. Conventional diagnostic techniques, which rely on the analysis of experts, are often laborious, expensive and less accessible, particularly in isolated regions. The present study proposes an automated plant disease detection system optimized using deep learning. The system utilizes Convolutional Neural Networks (CNNs). The proposed approach integrates advanced preprocessing techniques, including data augmentation, resizing, and normalization, to enhance model robustness and generalization. To facilitate deployment on mobile devices with limited resources, the MobileNetV2 architecture was optimized through quantization and conversion to TensorFlow Lite (TFLite). This approach resulted in a substantial reduction in computational complexity while maintaining an elevated level of classification accuracy. The mobile application, developed using Kotlin, facilitates the capture or upload of plant images and the execution of real-time disease detection directly on the device, thus obviating server communication. The experimental results demonstrate that the MobileNetV2 (Optimized) model achieved an accuracy of 99.48%, an F 1-score of 99%, and an AUC of 1.00, thus confirming its effectiveness for real-world agricultural applications. This study demonstrates the considerable potential of lightweight and efficient AI-driven solutions to transform the realm of plant disease detection, thereby rendering precision agriculture more accessible, particularly in resourceconstrained environments. © 2025 Elsevier B.V., All rights reserved.
  • Thumbnail Image
    Conference Object
    A Lightweight Mobile Deep Learning Framework for Real-Time Plant Disease Detection in Smart Agriculture
    (Institute of Electrical and Electronics Engineers Inc., 2025) Avcı, İsa; Koca, Murat; Khan, Yahya Zakrya
    It is imperative to detect plant diseases early to enhance agricultural productivity and ensure food security. Conventional diagnostic techniques, which rely on the analysis of experts, are often laborious, expensive and less accessible, particularly in isolated regions. The present study proposes an automated plant disease detection system optimized using deep learning. The system utilizes Convolutional Neural Networks (CNNs). The proposed approach integrates advanced preprocessing techniques, including data augmentation, resizing, and normalization, to enhance model robustness and generalization. To facilitate deployment on mobile devices with limited resources, the MobileNetV2 architecture was optimized through quantization and conversion to TensorFlow Lite (TFLite). This approach resulted in a substantial reduction in computational complexity while maintaining an elevated level of classification accuracy. The mobile application, developed using Kotlin, facilitates the capture or upload of plant images and the execution of real-time disease detection directly on the device, thus obviating server communication. The experimental results demonstrate that the MobileNetV2 (Optimized) model achieved an accuracy of 99.48%, an F 1-score of 99%, and an AUC of 1.00, thus confirming its effectiveness for real-world agricultural applications. This study demonstrates the considerable potential of lightweight and efficient AI-driven solutions to transform the realm of plant disease detection, thereby rendering precision agriculture more accessible, particularly in resourceconstrained environments. © 2025 Elsevier B.V., All rights reserved.