Health Detection of Medicinal Plants Vitex negundo and Ricinus communis Using Transfer Learning

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Plant diseases pose a significant challenge to global agriculture, with early diagnosis particularly problematic for resource-limited farmers. This study introduces an AI-powered web application for classifying leaf health in two key medicinal plants, Ricinus communis (Eranda) and Vitex negundo (Nirgundi). The tool categorizes leaf images into three health states—Healthy, Medium Healthy, and Unhealthy to facilitate timely and informed decision making in crop management. A dataset comprising 2834 images of Ricinus communis and 3796 images of Vitex negundo under various conditions was created and used to train the model. Advanced transfer learning architectures, including VGG16, ResNet50, MobileNetV2, Xception, EfficientNetB0, and VGG19, were employed to enhance the classification accuracy of the system. Notably, VGG16, EfficientNetB0, and VGG19 achieved highest level of accuracy, whereas the other models showed comparatively lower level of accuracies. This work performs real-time assessment of leaf health for sustainable Ayurvedic agriculture.