AyushNet – an IoT-based Mobile App for the Automatic Recognition of Medicinal Plants based on a deep residual neural network

Traditional medicinal herbs are increasingly being integrated into modern medicine by the pharmaceutical and healthcare industries due to their lower cost and fewer side effects. The classification of medicinal plants significantly impacts Traditional Indian Medicine (TIM) resource protection, authentication, and identification instruction. The global healthcare community has recently shown more interest in Indian medicinal plants and their therapeutic benefits. Consequently, most research focuses on automatically identifying medicinal plants using their leaf, flower, or other plant organs. This requires using DL techniques for robust recognition, as there is minimal variation in image texture and color. In this study, we introduce AyushNets, a self-supervised mobile application that eliminates manual classification and predicts the label of medicinal plants based on leaf images. The leaf images of medicinal plants are collected from mobile devices, and the preprocessed data is trained using the Inception-Resnet V2 architecture. The knowledge gained from the trained data is further enhanced through the Transfer Learning Modal with newly available data. AyushNet achieves an average success rate of 97% when analyzing leaf photos of 30 different medicinal plants. This outcome suggests that automatic classification of medicinal plants using leaf images is feasible. The study provides a valuable conceptual framework for investigating and developing a classification system for medicinal plants.