MPInet: Medicinal Plants Identification using Deep Learning

Abstract

Botanists, chemists, and healthcare professionals have substantial difficulties in classifying and identifying medicinal plants. The conventional techniques for identifying plants are hard, time-consuming, and need a great deal of knowledge. Convolutional neural networks (CNN), a deep learning approach, have paved the path for the automatic identification of medicinal plants based on leaf photographs. The proposal named MPInet is a VGG16 model trained on custom collected dataset specifically for identifying the 10 medicinal plants which grows commonly around populated areas. The datset has 3000 samples spread across 10 species. The proposed model achieved performance accuracy of 99.4 percent, precision value of 99.6 percent, recall value of 99.5 percent and F1 Score of 98.8 percent. MPInet can be integrated to mobile devices and may be used for educational purpose.