Abstract:
Medicinal leaves are traditionally widely used in Bangladesh. Which plays a vital role in
protecting the health of the human body. Bangladesh is a country where we find many plant
species and each plant has its medicinal properties. The manual identity and classification
process knowledge very difficult for human beings to remember all plant-specific names
and uses. These medicinal leaves are very important and beneficial to many sectors such
as the medical field, botanic research, and plant classification study. So, Medicinal leaves
of Bangladesh can preserve this resource by researching the classification of medicinal
leaves. Our research aims to accurately identify medicinal leaves by proposing an active
classification system based on medicinal leaves. Our method is to first correctly preprocess
the medicinal plant leaves then it will classify the leaves of the medicinal plant using
DenseNet201, ResNet50V2, and InceptionV3 models. The models have been applied to 10
different classes their total 2094 original medicinal plant leaf images. Where DenseNet201
provides the highest 96.06% accuracy. The result indicates that it is feasible to
automatically classify medicinal plants. The paper provides a valuable theoretical
framework for the research and development of the medicinal plant classification system
