Medicinal Plant Identification in Real-Time Using Deep Learning Model

• Published in: SN computer science Vol. 5; no. 1; p. 73
• Main Authors: Kavitha, S., Kumar, T. Satish, Naresh, E., Kalmani, Vijay H., Bamane, Kalyan Devappa, Pareek, Piyush Kumar
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.01.2024; Springer Nature B.V
• Subjects: Accuracy; Analytics and Networks; Applications programs; Automation; Ayurvedic medicine; Botany; Computer Imaging; Computer Science; Computer Systems Organization and Communication Networks; Data Structures and Information Theory; Deep learning; Diverse Applications in Computing; Flowers & plants; Herbal medicine; Herbs; Information Systems and Communication Service; Leaves; Machine vision; Medical research; Medicinal herbs; Mobile computing; Original Research; Pattern Recognition and Graphics; Pharmaceuticals; Plant diseases; Quality control; Real time; Software Engineering/Programming and Operating Systems; Vision; Borneo; Augmentation; Deep learning; Medicine; Plant; Accuracy; Resize; Cloud; Mobile application
• DOI: 10.1007/s42979-023-02398-5
• ISSN: 2661-8907; 2662-995X

Medicinal plants have a long tradition of being cultivated and harvested in India. The Indian Forest is the principal repository for many useful medicinal herbs. As a result of their critical role in maintaining people's life, medicinal plants have traditionally been the subject of intensive research and consideration. Yet, correctly identifying plants used in medicine is a laborious process that takes a lot of time and expertise. Because of this, a vision-based approach may aid scientists and regular people in the rapid and precise identification of herb plants. Therefore, this research suggests a vision-based smart method to recognize herb plants by creating a deep learning (DL) model. Although there is a wide variety of useful plants, we limit ourselves to just six from the Kaggle database: betel, curry, tulsi, mint, neem, and Indian beech. For each medicinal plant, we collected 500 images. The data undergo a process of resizing and augmentation to increase the sample size. For the fully automatic identification of medicinal leaves, the MobileNet DL model is selected. To determine the model's effectiveness, it must first be trained, then validated, and ultimately tested. The DL model is evaluated using measures including accuracy, precision, and recall. For this reason, the DL model was able to correctly identify medicinal leaves at an accuracy rate of 98.3%. After being thoroughly investigated, the DL model is uploaded to the cloud, and a mobile app is created for the real-time identification of medicinal leaves. To recognize leaf images, the built mobile app accesses the DL model on the cloud. The automated recognition of plants represents an extremely promising option for filling the taxonomic gap and gaining a lot of interest from the fields of botany and machine vision.