Automatic Identification and Revealing Pertinent Information AIRPI of Herbal Leaves

Abstract

Plants play a vibrant role in several expanses such as drug formulation, agriculture, pharmaceutical newlineindustry and in maintaining the ecological balance. Several plants possess newlinepotential medicinal properties that cure varied common ailments and diseases. Global newlinewarming, urbanization and many more human activities have resulted in the extinction newlineof many medicinal herbs. Moreover, the knowledge of medicinal plants is well known newlineto experts such as botanists and taxonomists. Compared to synthetic drugs (modern newlinemedicine), many developing and developed countries rely on the traditional system of newlinemedicine due to its reduced side effects and low cost. Ayurveda, a traditional medicinal newlinesystem of India, utilizes various medicinal plants and their different parts to cure chronic newlinediseases and common ailments. Depending on limited experts and manual recognition newlineof the herbs is both time-consuming and a tedious task. The recent surge in the domain newlineof artificial intelligence that consists of many sub-domains for instance machine newlinelearning, deep learning and computer vision focus to build intelligent systems to solve newlinevarious domain problems. As an alternative to manual identification, the current research newlinefocuses to amalgamate Ayurveda and computer science to develop an automatic newlinesystem to categorize the Indian plants with medicinal properties using varied deep learning newlineand machine learning techniques. The proposed research work develops a machine newlinelearning model by implementing feature extraction method such as Scale Invariant Feature newlineTransform (SIFT) and support vector machine (SVM), naive bayes and k-nearest newlineneighbour as classifiers. A deep learning model developed by introducing the transfer newlinelearning technique on pre-trained deep networks namely, VGG-19, VGG-16, Xception newlineand Inception-V3 architecture for the extraction of the features and the artificial neural newlinenetwork (ANN), SVM with Bayesian optimization technique and SVM for classification. newlineOf the many models developed, the proposed model entitled as AIRPI includes the techni