Auto-Detection of Medicinal Plants using Machine Learning Approach

Abstract

This study investigated the machine-learning algorithms TensorFlow Lite and PyTorch to automatically identify five plants with medicinal properties: Cimamlilo, Hlaka-Hlakana, Mhlonyane, Mpepho, and Mthoba. Increasing demand for herbal remedies and lack of knowledge regarding plant identification among individuals has motivated development of an automated detection model. The proposed solution combines computer vision, image processing, and machine learning models to accurately detect and classify five medicinal plants. Effectiveness of the application was evaluated by conducting a statistical analysis to evaluate the performance of TensorFlow and PyTorch in detecting these plants under different imaging angles at different times of the day. Overall performance indicated that PyTorch was the best model for classifying these plants, with a detection accuracy of 85% at an angle of 90°.