Detecting Medicinal Plants Using YOLOv5: A Mobile Vision Approach

The use of medicinal plants for treating ailments has been a practice since ancient times. With the advent of computing and technology, the field of automated plant species identification has rapidly advanced. While there have been many studies on classifying plant species using leaves, the automatic identification using other plant organs remains unexplored. In this paper, we aim to investigate the precise and automated detection of flowers from medicinal plants in their natural environment. To achieve this, a new medicinal plant object detection dataset is presented comprising of four classes of medicinal plant species commonly found in the Philippines. Then, trained a YOLOv5-based model that detects medicinal plant flowers with a mean average precision of 83%. And, a mobile application was developed with the integrated model. Overall, the findings demonstrate the feasibility and acceptability of real-time automated detection and identification of medicinal plant species. This research has significant implications for the future of traditional medicine and the preservation of medicinal plant species.