Scholarlink

The precise classification of the plant is presupposed in different fields of science, such as the preservation of biodiversity, agriculture, medicine, pharmacology. Although there are more than 400,000 known plant species, there are still issues with identity and classification, because of morphological similarities and intra-environmental variations of plants. Recent development in the field of deep learning has led to a big boost in automated recognition of plants especially the use of leaf images which is popular because of its availability and unique characteristics. The work studies the complementary use of deep learning in feature extraction and meta-heuristic optimization in order to boost the classification accuracy. Namely, it considers the performance of using a VGG16 model pre-trained in deep features extraction that is associated with a Binary Particle Swarm Optimization (BPSO) to select the features. A Support Vector Machine (SVM) classifier was used to arrive at the end classification. Three benchmark datasets, MalayaKew, PlantVillage, and Swedish Leaf, were used as benchmarks in the experimentation. The suggested method was able to achieve 92.3%, 99.4%, and 94.3 percent classification accuracies among various plant species proving to be robust and efficient.

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

Plants Leaves Classification Using VGG16-Based Transfer Learning and PSO Algorithm

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Zainab Falah Jaber Abed,
  • Saeed M. Hashim

摘要

The precise classification of the plant is presupposed in different fields of science, such as the preservation of biodiversity, agriculture, medicine, pharmacology. Although there are more than 400,000 known plant species, there are still issues with identity and classification, because of morphological similarities and intra-environmental variations of plants. Recent development in the field of deep learning has led to a big boost in automated recognition of plants especially the use of leaf images which is popular because of its availability and unique characteristics. The work studies the complementary use of deep learning in feature extraction and meta-heuristic optimization in order to boost the classification accuracy. Namely, it considers the performance of using a VGG16 model pre-trained in deep features extraction that is associated with a Binary Particle Swarm Optimization (BPSO) to select the features. A Support Vector Machine (SVM) classifier was used to arrive at the end classification. Three benchmark datasets, MalayaKew, PlantVillage, and Swedish Leaf, were used as benchmarks in the experimentation. The suggested method was able to achieve 92.3%, 99.4%, and 94.3 percent classification accuracies among various plant species proving to be robust and efficient.