摘要
中草药是传统医学的重要组成部分,但分类效率低。为此,采用深度学习技术自动分类中草药图像。在公开的中草药数据集上评估了VGG、ResNet、Inception和DenseNet等模型,并建立了性能基线。通过决策级融合策略,对这四个模型的预测结果进行加权集成,显著提高了分类准确率。实验结果显示,集成模型的测试集准确率达到96.91%,明显优于任何单一模型。这不仅提高了中草药图像的分类效率和准确性,还为深度学习技术在传统医学领域的应用开辟了新途径。
Traditional Chinese herbs are an integral part of traditional medicine,but the classification efficiency is low.To address this,this study utilized deep learning technology for automatic classification of Chinese herbal images.Several models including VGG,ResNet,Inception,and DenseNet were evaluated on a publicly available Chinese herb dataset,establishing a performance baseline.A decision‑level fusion strategy was employed to weight the ensemble of predictions from these four models,significantly enhancing classification accuracy.Experimental results show that the ensemble model achieved a test set accuracy of 96.91%,substantially surpassing any individual model.This not only improves the efficiency and accuracy of Chinese herbal image classification but also paves new avenues for the application of deep learning technology in the field of traditional medicine.
作者
张超辉
Zhang Chaohui(Center for Technology Enhanced Learning,Guangdong Maoming Health Vocational College,Maoming 525400,China)
出处
《现代计算机》
2024年第18期99-103,共5页
Modern Computer
基金
广东省普通高校特色创新项目(2021KTSCX326)。
关键词
中草药
图像分类
深度学习
模型集成
Chinese herbal medicine
image classification
deep learning
model ensemble
