Identification of Fake Medicinal Raw Materials using Machine Learning Techniques

This paper extends the methodology for leaf classification by utilizing ResNet50, particle swarm optimization for hyperparameter tuning (PSO), and support vector machines (SVM) to classify 30 species, compared to the previous work that classified 7 species. The proposed approach pulls down the powerful feature extraction capabilities of ResNet50, optimized through PSO for hyperparameter tuning, and classified with SVM. By incorporating these advanced techniques, our approach achieves a high accuracy of 99.51%, which significantly outperforms previous methods. The scalability and robustness of this combined model are thoroughly demonstrated through extensive experimentation and evaluation on a comprehensive dataset of over 3500 leaf images. The results highlight significant improvements in classification performance, showing the model’s effectiveness in handling a larger and more diverse set of species. This advancement underscores the potential for broader applications in various fields, including botanical studies, agricultural automation, and environmental monitoring. By providing a scalable solution that maintains high accuracy, our methodology can be utilized for large-scale plant identification and monitoring systems, contributing to more efficient agricultural practices and better understanding of plant biodiversity. The promising results set a new benchmark for research and development in the domain of leaf classification in future, encouraging further exploration of deep learning techniques and optimization methods to enrich classification accuracy and applicability across different plant species and environmental conditions.