Revitalizing Ayurveda with Deep Learning: Automated Identification of Medicinal plants using CNN
Dr. Shaik Mahammad Rafee1 ,N. Deepak Sai2, K. Pravallika3, K. Varshini4, M. Ajay5
Department of Artificial Intelligence and Machine Learning
Sasi Institute of Technology and Engineering, Tadepalligudem, Andhra Pradesh, India mohammadrafee@sasi.ac.in, deepak.nambula@sasi.ac.in , pravallika.kodakalla@sasi.ac.in varshini.karuturi@sasi.ac.in , ajay.mallavarapu@sasi.ac.in
ABSTRACT: From ancient times ,Indians are used to grow and use medicinal herbs, and they have been an essential part of human life for ages. Indian forests are home to different medicinal plants, which have become key subject of scientific exploration due to their vital role in health and healing. But correctly recognizing these species is still a difficult and time- consuming procedure. This study proposes an intelligent vision-based method using deep learning (DL) to identify medicinal plants.500 photos of each of the six plant species—Indian beech, curry, tulsi, mint, neem, and betel—were gathered from a various sources and different existing dataset. To improve its quality, the dataset was preprocessed using augmentation and scaling techniques. We used the MobileNet DL model, which went through extensive training, validation, and testing stages, for the fully automated identification of medicinal leaves. The model's efficacy was assessed based on key performance metrics including
,accuracy, precision, and recall. The models accuracy is 98.3% which is a makeable outcome.After extensive training the model is deployed into a mobile application. This application uses a cloud-based tool for leaf image processing, which gives immediate identification results. With extensive applications in botany, the natural sciences, and computer vision, the automatic plant identification system addresses a legitimate requirement in the identification of medicinal plants.
KEYWORD: Convolution neural network, KNN, Gaussian Mixture model, Image Processing, and Python.
