Identification of Medicinal Herbs using Modified Xception Architecture

Identifying medicinal herbs is important for traditional medicine, pharmacology, and protecting biodiversity. This paper proposes a novel way of identifying medicinal herbs from an image dataset. Different deep learning methods have been performed, including using pre-trained neural networks and creating custom models for herb recognition. A comprehensive analysis of convolutional neural network (CNN) models for medical picture recognition tasks, with an emphasis on architectures like VGG16, ResNet50, and several Xception versions has been done. After in-depth analysis, it can be concluded that the modified Xception architecture performs significantly better than the others, attaining the highest training and validation accuracies with the least amount of losses. The results provide important implications for healthcare applications by highlighting the crucial role that model design plays in improving the accuracy of medicinal image analysis. It also describes how AI-driven tools can help better identify medicinal herbs from images, leading to significant progress in the field of traditional medicines and pharmacology.