Deep learning-based medicinal plant identification using hybrid ConvNeXt-Large and RegNetY-3.2GF feature fusion

Le résumé fourni par la source

Medicinal plants play an important role in traditional and modern healthcare, offering natural remedies for a wide range of diseases. However, accurately identifying medicinal leaves remains challenging due to high interspecific similarity. Various existing studies have explored deep learning models for plant and leaf identification, but there is still scope for improvement in accuracy and efficiency. We have proposed a hybrid deep learning model that combines ConvNeXt-Large and RegNetY-3.2GF through feature fusion, leveraging the complementary strengths of both architectures. Multiple convolutional neural network (CNN) models have been evaluated on a medicinal plant leaf dataset, employing five-fold cross-validation and assessing performance using metrics including accuracy, precision, recall, F1-score, and ROC-AUC. The results show that our hybrid approach outperforms each individual architecture, achieving a mean classification accuracy of 98.79%, a precision of 98.015%, a recall of 99.68%, and an F1-score of 98.59%, with minimal standard deviation across folds. This research highlights the potential of deep learning in medicinal plant identification, which will be helpful for researchers and practitioners in this field.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.