PJB-2025-432
CC-MobileViT: A Lightweight Model with Mobile Application for Apple Leaf Disease Automated Recognition
Mingjian Zhang
Abstract
The cultivation of apple trees plays an irreplaceable strategic role within the world’s agricultural system. To mitigate the damaging effects of apple leaf diseases on crop production and food security, it is imperative to implement deep learning (DL)-driven automated systems for early diagnosis. Conventional DL models generally demand designing deep and extensive structures aimed at improving feature extraction performance, but this will lead to the cost of computational overhead. To address this problem, a lightweight leaf disease recognition model called CC-MobileViT is proposed in this paper. CC-MobileViT is an enhanced version of MobileViT, in which the standard convolution is superseded by the CompConv. Compared to MobileViT, the CC-MobileViT model achieves a 40% reduction in parameter count while maintaining the high detection accuracy of 98.10%, which outperforms the state-of-the-art DL models. Moreover, the model has been deployed on the mobile-based WeChat Mini Program, which provides farmers and researchers with the convenience of capturing or uploading images from their smartphones for real-time detection. Therefore, the proposed model strikes a favorable balance between the efficiency and accuracy of disease diagnosis.
To Cite this article:
Under Review
We Welcome Latest Research Articles In Field Of Botany The Pakistan Journal of Botany is an international journal .... Read More