Implementing an efficient image processing technique with deep neural networks for the classification of medicinal herbs

Abstract

Plants, in all their diversity, represent a valuable contribution from the environment. When illness strikes, whether physical or mental, medicine becomes our primary recourse. Many pharmaceuticals are derived from natural plants, which serve as vital sources of medicinal compounds. In Bangladesh, these medicinal plants are also known by the names Homeopathy, Unani, and Ayurveda. Experts suggest that the COVID-19 pandemic can be effectively addressed with the aid of medicinal plants. Strengthening our immune system is paramount, as it directly influences overall health. A robust immune system can combat bacteria, viruses, and other pathogens. Conversely, inactive individuals are more susceptible to viral infections and diseases. Certain medicinal plants have been shown to enhance immunity. Therefore, accurately classifying these plants is crucial. We proposed using four well-known algorithms—DenseNet201, VGG19, and ResNet152—to classify medicinal plants based on leaf images. Our approach achieved accuracies of 96.08% with DenseNet201, 97% with VGG19, 98% with ResNet152, and an impressive 99.12% with a hybrid model combining VGG19 and ResNet50. These results highlight the potential of advanced deep learning techniques in enhancing the identification and utilization of medicinal plants. Our research will help people understand their immune systems better and the medicinal plants that can boost immunity, thereby enabling them to combat diseases and viruses more effectively in the future. As we look ahead, it is important to acknowledge that dangerous infectious viruses may continue to emerge. Our study not only contributes to the field of medicinal plant classification but also underscores the importance of integrating traditional medicinal knowledge with modern technological advancements to improve public health outcomes.