Citation

Hossain, Monir and Farid, Fahmid Al and Begum, Momotaz and Uddin, Jia and Karim, Hezerul Abdul (2026) Deep Learning for Medicinal Plant Classification: A Comprehensive Review of Recent Advances and Challenges. Journal of Computational and Cognitive Engineering, 5 (3). pp. 544-560. ISSN 2810-9570

Text Deep Learning for Medicinal Plant Classification_ A Comprehensive Review of Recent Advances and Challenges.pdf - Published Version Restricted to Repository staff only Download (381kB)

Abstract

Deep learning classifies medicinal plants, driven by the need to preserve traditional knowledge and automate identification for practical uses. This review extensively summarizes 30 recent studies (2021–June 2025) on applying deep learning, primarily using image data, to classify medicinal plants. This review analyzes research distribution, dataset preparation, image preprocessing, augmentation, and deep learning architectures like convolutional neural networks, Vision Transformers, and hybrid models. Our analysis reveals a strong geographic focus, with 50% of the selected studies originating from India and Bangladesh. The focus is overwhelmingly on leaf imagery, with 29 out of the 30 studies relying on this approach. The field is also characterized by its dependence on existing data, as 56.6% of studies utilized public datasets and another 26.6% employed a hybrid of public and private data, with dataset sizes ranging from a minimum of 637 to a maximum of 13,500 images. Methodologically, the vast majority of studies rely on a transfer learning approach (36.7%), achieving robust accuracy rates between 74% and 99.9%. Furthermore, we recognize significant limitations, such as the absence of standardized and diverse datasets, insufficient inclusion of uncommon or endangered species, and inadequate representation of whole-plant imaging. The research underscores the necessity for collaborative, multidisciplinary initiatives to develop centralized, high-quality, and geographically comprehensive datasets. We delineate prospective avenues, including multimodal feature integration, the development of real-world applications, and optimization for privacy-preserving frameworks such as federated learning. This study guides academics advancing deep learning for medicinal plant classification and biodiversity conservation.

Item Type:	Article
Uncontrolled Keywords:	Plant image dataset, deep learning
Subjects:	Q Science > QA Mathematics > QA71-90 Instruments and machines > QA75.5-76.95 Electronic computers. Computer science
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	02 Oct 2026 00:37
Last Modified:	02 Oct 2026 00:37
URII:	http://shdl.mmu.edu.my/id/eprint/16808

Downloads

Edit (login required)