Identification and Classification of Medicinal Leaves and their Medicinal Values

The traditional knowledge systems about medicinal properties of plant species have started to disappear with the passage of time due to the scarcity of written documents. Due to resemblance in the visual properties among plants, sorting and classification are complicated to carry out, especially in large area. The main objective of this work is to develop a machine learning based system for classifying different Indian medicinal plants, which will be useful in drug discovery process and provide a platform for future research. This work on identifying plant species and determining their medicinal value using the VGG-16 convolutional neural network model. We embarked on the task of collecting a comprehensive dataset by conducting fieldwork. The work involved visiting botanical gardens to collect plant specimens along with their associated information, such as species, habitat, and traditional medicinal uses. The collected images were labelled ensuring its quality and relevance for training our model. The collected dataset was then utilized to train the VGG-16 model. Techniques such as data augmentation and transfer learning to enhance the model’s performance and optimize training efficiency. Throughout the training process the model’s parameters and hyper parameters were fine-tuned, striving to achieve optimal accuracy and robustness. After conducting extensive research and comparison between VGG 16, Resnet 50, Sequential Model. VGG-16 yielded superior results based on our findings. The VGG-16 model achieved an impressive accuracy of 98%, outperforming other models.