Scholarlink

<p>Medicinal plants have been utilized since ancient times. However, with the advancement of modern medicine, their use diminished. Due to the adverse effects associated with allopathic treatments, there has been a resurgence of interest in herbal medicine. This study investigates the classification of herbal plant leaves using deep learning techniques, specifically convolutional neural networks (CNNs). Experiments were conducted using a dataset of 1,835 images from 30 different herbal plant species, applying five pre-trained models: VGG16, VGG19, ResNet50V2, MobileNetV2, and InceptionV3. Among these models, ResNet50V2 emerged as the best performer, achieving an accuracy of 96%. Even MobileNetV2 (95%) and InceptionV3 (92%) produced commendable results that are suitable for lightweight applications. However, VGG16 and VGG19 appeared to generalize less effectively on the dataset, with lower accuracy rates of 86% and 82%, respectively. Thus, it can be concluded that deeper and more optimized architectures are better suited for the task of herbal plant classification. The study demonstrates how deep learning can automate the classification of herbal plants, providing significant benefits to agriculture, medicine, and biodiversity.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep learning models for herbal plant leaf classification: a comparative analysis

  • 期刊论文

  • 发表日期：

    2026年1月29日

查看全文
  • Megha Raina,
  • Umar Bashir,
  • Vibhakar Mansotra

摘要

Medicinal plants have been utilized since ancient times. However, with the advancement of modern medicine, their use diminished. Due to the adverse effects associated with allopathic treatments, there has been a resurgence of interest in herbal medicine. This study investigates the classification of herbal plant leaves using deep learning techniques, specifically convolutional neural networks (CNNs). Experiments were conducted using a dataset of 1,835 images from 30 different herbal plant species, applying five pre-trained models: VGG16, VGG19, ResNet50V2, MobileNetV2, and InceptionV3. Among these models, ResNet50V2 emerged as the best performer, achieving an accuracy of 96%. Even MobileNetV2 (95%) and InceptionV3 (92%) produced commendable results that are suitable for lightweight applications. However, VGG16 and VGG19 appeared to generalize less effectively on the dataset, with lower accuracy rates of 86% and 82%, respectively. Thus, it can be concluded that deeper and more optimized architectures are better suited for the task of herbal plant classification. The study demonstrates how deep learning can automate the classification of herbal plants, providing significant benefits to agriculture, medicine, and biodiversity.