Dataset of Selected Medicinal Plant Species of the Genus Brachylaena: A Comparative Application of Deep Learning Models for Plant Leaf Recognition
Résumé
Since several active pharmaceutical ingredients are sourced from
medicinal plants, identifying and classifying these plants are generally a valuable
and essential task during the drug manufacturing process. For many years,
identifying and classifying those plants have been exclusively done by experts in
the domain, such as botanists and herbarium curators. Recently, powerful
computer vision technologies, using deep learning or deep artificial neural
networks, have been developed for classifying or identifying objects using
images. A convolutional neural network is a deep learning architecture that
outperforms previous state-of-the-art approaches in image classification and
object detection based on its efficient feature extraction of images. This study
investigated several pre-trained convolutional neural networks for identifying and
classifying leaves of three species of the genus Brachylaena. The three species
considered were Brachylaena discolor, Brachylaena ilicifolia, and Brachylaena
elliptica. All three species are used medicinally by people in South Africa. We
trained and evaluated different deep convolutional neural networks from 1259
labeled images of those plant species (at least 400 for each species) split into
training, evaluation, and test sets. The best model provided a 98.26% accuracy
using cross-validation with a confidence interval of ±2.16%.
Origine : Fichiers éditeurs autorisés sur une archive ouverte