부산대학교 도서관

Medicinal plants are useful in various disease conditions. The Medicinal plant contains many bioactive vital components that are used to prepare different varieties of Ayurvedic formulations. The Medicinal plants, with their medici-nal value, preparation methodology, and identification techniques are mentioned in ancient foundational texts like Charak Samhita. However, most of the identification techniques a re subjective. Also with a global surge in Ayurveda remedies, the common man is also interested in remedies using Ayurvedic plants. Hence we used AI for the identification o f medicinal plants. Using a customized dataset of 95 medicinal plant species, we trained SSD MobileNet v2 for medicinal plant species detection (with 87.22% Precision and 89.79% Recall) and VGG16 for medicinal plant species classification (with 9 5 % accuracy). Furthermore, primary leaf morphological features - venation of the leaves (with 98.23% accuracy), and margin or edges of the leaves (with 98.83% accuracy), are also extracted and used to make an accurate recommendation.