Publication Type

Journal Article

Version

acceptedVersion

Publication Date

11-2025

Abstract

Accurately identifying crop diseases plays a crucial role in advancing intelligent and modern agricultural production. Deep learning techniques have performed robust performance in classifying plant disease images. However, current studies face the challenge that many plant disease datasets are generated in controlled environments, leading to reduced model performance in real-world agricultural settings. This paper aims to provide a lightweight model that can accurately classify plant diseases in natural environments. Specifically, this paper investigates the Dual-Attention Multi-Scale Lightweight Network (DAMSLNet), which combines dual-attention-based multi-scale feature extraction and deep information fusion, to classify plant diseases. At the front end, the model employs depth-wise separable convolution (DSC) to extract shallow features from images, followed by the improved InceptRes module for multi-scale deep feature extraction. The back end further focuses the network's attention on the diseased part through the dual attention module (DA Module), while ignoring unimportant features. The proposed DAMSLNet has been trained and verified on four datasets, achieving a classification accuracy of 98.19%, 99.96%, 97.08%, 99.99% respectively. The complexity of the DAMSLNet is about 16% of InceptionV4, and the parameter volume is also reduced to 37.5% of InceptionV4. In addition, this paper used technologies such as Gradient-weighted Class Activation Mapping (GradCAM) to perform interpretability analysis on the performance of different modules in the network. The DAMSLNet was compared with state-of-the-art disease classification networks. The experimental results show that our proposed network has better classification performance compared to existing advanced methods.

Keywords

Deep learning, Disease classification, Dual attention, Interpretability analysis, PlantVillage, Xinong Apple dataset

Discipline

Databases and Information Systems | Plant Sciences

Research Areas

Data Science and Engineering

Publication

Engineering Applications of Artificial Intelligence

Volume

159

First Page

1

Last Page

16

ISSN

0952-1976

Identifier

10.1016/j.engappai.2025.111590

Publisher

Elsevier

Copyright Owner and License

Authors

Additional URL

https://doi.org/10.1016/j.engappai.2025.111590

Download

Share

COinS