Herb Detect: Medicinal Plant Identification System

The importance of medicinal plants is increasingly appreciated owing to the roles they play in traditional as well as modern medicine. However, identification of these medicinal plants is often challenged by the visual resemblances that certain non-medicinal species have with them. The contribution of this project is to develop a Medicinal Plant Classification System which employs Deep Learning in automation of the identification process. The One-Class Autoencoder, a modified segregated model is trained to identify medicinal and non- medicinal plants, guaranteeing a certain assurance about the correctness of the classification. In addition, a Convolutional Neural Network (CNN) based on MobileNetV2 identifies specific medicinal plant species correctly. The trained model is deployed as a web application using Flask in which any user uploads the image of a plant and receives a prediction instantly with descriptions of its medicinal properties if the plant is medicinal. This system will support botanists, researchers, and farmers with the accurate identification of medicinal plants in their effort towards conservation and sustainable use. Key Words:Plant classification, MobileNetV2, Deep Learn- ing, Image identification.