Identification of Medicinal Leaves Using State of Art Deep Learning Techniques

India has a long history of being a great source of medicinal herbs. In India, the forest is the primary storehouse for a significant number of medicinal plants. This paper demonstrates robustness of deep neural networks model for automatically identifying different kinds of medicinal leaves using state-of-art deep learning techniques with transfer learning models. The majority of people nowadays are unaware of their local medicinal plants and their benefits in both rural and urban areas. In addition, manual recognition of different kinds of leaves is challenging. So, in that context applied three transfer learning techniques VGG16, MobileNet and InceptionResNetV2 for CNN model to classify the medicinal leaves in reduced processing time. The developed mobile app identifies the medicinal leaf by capturing medicinal images on the fly either by mobile or by selecting image from data repository. The app also provides additional information such as botanical name, Indian name and uses of the leaf identified. Mobile Application predicts the live prediction also by take image of the leaf with the help of camera and white background only. To the best of our knowledge, no work is carried on the publicly available medicinal leaf dataset used in this paper with transfer learning.