ORIGINAL RESEARCH article
Front. Plant Sci.
Sec. Sustainable and Intelligent Phytoprotection
Volume 16 - 2025 | doi: 10.3389/fpls.2025.1656731
This article is part of the Research TopicCutting-Edge Technologies Applications in Intelligent Phytoprotection: From Precision Weed and Pest Detection to Variable Fertilization TechnologiesView all 13 articles
LDL-MobileNetV3S: An Enhanced Lightweight MobileNetV3-Small Model for Potato Leaf Disease Diagnosis through Multi-Module Fusion
Provisionally accepted
- Inner Mongolia Agricultural University, Hohhot, China
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
The timely and precise detection of foliar diseases in potatoes, a food crop of worldwide importance, is essential to agricultural output. In complex field environments, traditional recognition methods encounter three primary challenges: (1) the small size and morphological diversity of early-stage lesions pose significant difficulties for feature extraction; (2) the gradual transition between diseased and healthy tissues leads to blurred edge features; and (3) background interference from leaf texture and varying illumination substantially degrades recognition robustness. To address these challenges, this study proposes an optimized lightweight convolutional neural network architecture, termed LDL-MobileNetV3S. The proposed model builds upon the MobileNetV3 Small architecture and achieves performance improvements through a three-stage innovative design. First, a Lightweight Multi-scale Lite Fusion (LF) module is introduced to enhance multi-scale perception of small lesions via cross-layer connections. Second, a Dynamic Dilated Convolution (DDC) module employs deformable convolutions to adaptively capture pathological features with blurred boundaries. Finally, a Lightweight Attention (LA) module is incorporated to suppress background interference by assigning spatially adaptive weights. Experimental results indicate that the proposed model achieves a recognition accuracy of 94.89%, with corresponding Precision, Recall, and F1-score values of 93.54%, 92.53%, and 92.77%, respectively. These results are obtained under a highly compact configuration, with the model occupying only 6.17 MB of storage and comprising 1.50 MB of parameters substantially smaller than those of EfficientNet-B0 (15.61 MB / 3.83 MB) and ConvNeXt Tiny (106 MB / 27.8 MB). Overall, the proposed approach demonstrates superior performance compared to existing lightweight models. This study offers a cost-effective and high-accuracy solution suitable for intelligent diagnostic devices deployed in resource-limited field environments.
Keywords: potato leaf disease, MobileNetV3 Small, Lite Fusion, Dynamic Dilated Convolution, Lightweight Attention
Received: 30 Jun 2025; Accepted: 02 Oct 2025.
Copyright: © 2025 Zhang, Yang, Fu, Wang and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Honghui Li, lihh@imau.edu.cn
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.