A Robust Deep Learning Framework for Automated Medicinal Plant Identification Through Leaf Image Analysis

The primary objective of this research is to develop a robust Medicinal Plant Classification System by leveraging the Deep Learning concept. Accurate identification and classification of these plants are critical for advancing pharmacological applications and ensuring effective treatment protocols. The system was built upon a proprietary dataset comprising 1,500 images gathered across 30 distinct Indian medicinal plant species, with each species represented by approximately 50 to 60 images. The feature set utilized for classification incorporates the plants’ leaf texture, shape, color, and key physiological/morphological characteristics. A Convolutional Neural Network (CNN) architecture was employed to train the data and achieve a highly accurate classification model. Through training several model architectures, the best-performing iteration yielded a 96.67% success rate in identifying the correct medicinal plant species. This significantly high accuracy validates the model’s potential as a valuable advisory and early warning tool for quick and reliable identification. .