Automated Medicinal Plant Recognition Using Capsule Networks

Medicinal plants are necessary for traditional as well as modern medicine, offering natural medication for a wide range of conditions. Accurate classification of these plants is essential to ensure their correct identification, which is crucial for safe usage, drug development, and conservation efforts. Misidentification can lead to ineffective treatment or even harmful effects. Additionally, proper classification aids in preserving biodiversity, supporting sustainable use, and protecting endangered species. The proposed system offers an innovative approach to medicinal plant identification by integrating Capsule Networks and the YOLO algorithm, addressing the critical need for accuracy in traditional medicine, conservation, and pharmacology. Manual identification is labor-intensive and error-prone, driving the development of automated systems. Conventional CNNs, while effective in classification, struggle with generalization and feature learning, and YOLO's use of rectangular bounding boxes can limit its accuracy in capturing complex leaf shapes. Capsule Networks maintain spatial hierarchies and handle affine transformations, enhancing feature extraction, while YOLO ensures real-time detection and accurate localization. The system was trained on a massive dataset of medicinal plant pictures, using upscaling and normalization techniques to improve precision. By leveraging the strengths of both Capsule Networks and YOLO, the architecture achieves superior performance, making it suitable for practical applications, including mobile devices.