Corn or maize leaf disease classification based on CNN and vision mamba model.

Abstract

Accurate classification of corn leaf diseases is critical for timely detection and control of pests and diseases. By accurately recognizing different types of leaf diseases, farmers and agricultural experts can quickly take targeted control measures to reduce crop losses and safeguard corn yield and quality. Since these corn leaf disease images usually contain complex backgrounds, similar lesion features, and limited labeling data, it causes traditional convolutional neural networks (CNNs) to easily confuse the lesion region with the background, making it difficult to distinguish between different disease types. To address these limitations, we propose G-ResNet, a hybrid CNN-Vision Mamba network that enhances disease-relevant feature learning through a hierarchical feature attention module and a scale feature attention module. It was demonstrated experimentally that G-ResNet can better classify maize leaf disease images. The code is available at https://github.com/gustafmy/g_resnet.git.

Plant phenotyping relevance

トウモロコシ葉の病害状態を画像から分類するCNN・Vision Mamba手法の開発と実験評価が研究の中心であり、植物病害フェノタイピングに該当する。

Code and data availability

The paper's analysis code (G-ResNet) is publicly available on GitHub with explicit availability statements, and the plant image dataset used for the disease classification experiments is a public Kaggle dataset explicitly cited in the Experiment section. The underlying data availability statement also mentions request,