一种新的轻量级混合CNN-ViT用于玉米叶病疾病分类的新型混合动力
Saber Mehdipour1, Seyed Abolghasem Mirroshandel2, Seyed Amirhossein Tabatabaei3
1Department of Computer Engineering, University of Guilan, Rasht, Guilan, Iran.
Scientific reports
|February 25, 2026
概括
这项研究引入了一种新的混合人工智能模型来诊断玉米植物疾病,达到99.90%的准确性. 这种自动化方法为农业提供了高效和精确的疾病检测.
科学领域:
- 农业科学 农业科学
- 计算机视觉 计算机视觉
- 人工智能的人工智能
背景情况:
- 玉米是一个关键的全球作物,由于植物疾病,面临着巨大的产量损失.
- 手动疾病诊断是劳动密集型,易出错,缺乏可扩展性.
- 使用深度学习的自动化方法显示出潜力,但在CNN和ViT等现有架构中面临局限性.
研究的目的:
- 开发一种轻量级的混合深度学习模型,用于准确高效地诊断玉米疾病.
- 克服卷积神经网络 (CNN) 在捕捉全球背景和视觉转换器 (ViT) 在数据和计算要求方面的局限性.
主要方法:
- 提出了一种由专家混合 (MoE) 架构启发的新型混合模型,集成CNN和ViT组件.
- 该模型根据输入图像特征,以适应性的方式平衡本地和全球特征提取.
- 在一个新的,现实世界的数据集上对模型进行了评估,其中包括完整的玉米植物图像.
主要成果:
- 在玉米疾病数据集上达到99.90%的异常高分类准确度.
- 显著超过了包括MobileViT,PiT,EdgeNeXt和DeiT在内的已有的最先进的模型.
- 证明了轻量化混合架构在植物疾病诊断中的有效性.
结论:
- 轻量化混合深度学习模型为高性能,自动化植物疾病诊断提供了可行的解决方案.
- 开发的模型显示出在农业环境中实际部署的巨大希望,以改善作物管理.
- 这种方法解决了对高效准确诊断工具的需求,以保障全球粮食安全.
更多相关视频
06:11Author Spotlight: Improved Methods for Preparing Transverse Sections and Unrolled Whole Mounts of Maize Leaf Primordia for Fluorescence and Confocal Imaging
Published on: September 22, 2023
4.3K
06:21Micron-scale Phenotyping Techniques of Maize Vascular Bundles Based on X-ray Microcomputed Tomography
Published on: October 9, 2018
9.4K
相关概念视频
Light Acquisition
9.7K
In order to produce glucose, plants need to capture sufficient light energy. Many modern plants have evolved leaves specialized for light acquisition. Leaves can be only millimeters in width or tens of meters wide, depending on the environment. Due to competition for sunlight, evolution has driven the evolution of increasingly larger leaves and taller plants, to avoid shading by their neighbors with contaminant elaboration of root architecture and mechanisms to transport water and nutrients.
9.7K
Classification of Illness
9.1K
The meaning of illness is individualized to each person who experiences an alteration in health. In contrast, disease is a medical term indicating a pathological change in the structure and function of the body or mind. It is a condition that has specific symptoms and boundaries.
An illness is a response to a disease in which the person's level of functioning is changed compared with a previous level. The general classification of illness includes acute and chronic.
Acute illness is severe...
An illness is a response to a disease in which the person's level of functioning is changed compared with a previous level. The general classification of illness includes acute and chronic.
Acute illness is severe...
9.1K
