EUVIMEDEUVIMED entdecken
Crossref

Recognition of Rice Leaf Diseases in MobileNetV3 Based on Dual‐Path Efficient Channel Attention and Multi‐Feature Fusion

Jianian Li, Long Gao, Xiaocheng Wang, Jiaoli Fang, Yang Luo, Zhu Yuan, Shaomin Chen

Concurrency and Computation: Practice and Experience · 2026-07 · 38 (14)

DOI: 10.1002/cpe.70825 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Concurrency and Computation: Practice and Experience, online erschienen am 14.07.2026. Verfasst von Jianian Li und 6 weiteren Autor:innen. Die Arbeit nennt 29 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Rice, as one of the most crucial global food crops, is highly susceptible to diseases during cultivation, making accurate disease identification essential for effective control. Current high‐precision diagnostic methods require substantial computational resources, limiting their practical application. This study proposed an improved MobileNetV3‐Large (MV3L) model for identifying three common rice leaf diseases (rice blast, brown spot, and leaf blight) along with healthy leaves. A dual‐channel efficient channel attention (ECA) structure integrated with global maximum pooling (GMP) was designed, replacing the original Squeeze‐and‐Excitation (SE) module in MV3L with an enhanced ECA module. Additionally, a lightweight multi‐feature fusion module was developed to aggregate shallow and deep features. These improvements reduced the number of model parameters while enhancing discrimination capability for visually similar blast and brown spot lesions. The dataset, comprising field images and Kaggle platform data, covered four categories: leaf blight, brown spot, blast, and healthy leaves. Experimental results demonstrated that the proposed model achieved 97.57% accuracy (Accuracy) with only 2.774 M parameters (Params) outperforming the original MV3L by 3.09% in average recognition accuracy. Notably, F1‐scores for rice blast and brown spot classification increased by 5.9% and 4.93%, respectively, while parameters were reduced by 31%. Compared to the best‐performing mainstream lightweight model MobileViT, improved MV3L exhibited 1.99% higher accuracy (Accuracy) with 56% fewer parameters (Params). The experimental evidence demonstrates a favorable trade‐off between diagnostic accuracy and parameter efficiency, suggesting the potential of the proposed model for rice disease identification under resource‐constrained agricultural scenarios.

Zitieren

APA

Li, J., Gao, L., Wang, X., Fang, J., Luo, Y., Yuan, Z., & Chen, S. (2026). Recognition of Rice Leaf Diseases in MobileNetV3 Based on Dual‐Path Efficient Channel Attention and Multi‐Feature Fusion. Concurrency and Computation: Practice and Experience, 38(14). https://doi.org/10.1002/cpe.70825

Vancouver

Li J, Gao L, Wang X, Fang J, Luo Y, Yuan Z, et al. Recognition of Rice Leaf Diseases in MobileNetV3 Based on Dual‐Path Efficient Channel Attention and Multi‐Feature Fusion. Concurrency and Computation: Practice and Experience. 2026;38(14). doi:10.1002/cpe.70825

BibTeX
@article{Li2026,
  author = {Li, Jianian and Gao, Long and Wang, Xiaocheng and Fang, Jiaoli and Luo, Yang and Yuan, Zhu and Chen, Shaomin},
  title = {Recognition of Rice Leaf Diseases in MobileNetV3 Based on Dual‐Path Efficient Channel Attention and Multi‐Feature Fusion},
  journal = {Concurrency and Computation: Practice and Experience},
  year = {2026},
  volume = {38},
  number = {14},
  doi = {10.1002/cpe.70825}
}
Kostenlos starten →