Deep learning for medicinal plant species classification and recognition: a systematic review

Area of Science:

• Botany and Computer Science
• Focus on biodiversity conservation and technological applications.

Background:

• Medicinal plant classification is vital for conservation but expert identification is challenging.
• Deep learning offers automated solutions for recognizing medicinal plant species.

Purpose of the Study:

• To systematically review research on deep learning for medicinal plant classification and recognition.
• To identify trends, methodologies, and research gaps in this field.

Main Methods:

• Conducted a systematic review of studies published between January 2018 and December 2022.
• Screened 1644 studies, selecting 31 for in-depth analysis following PRISMA guidelines.

Main Results:

• India leads research (29%), with most studies using private datasets and plant leaf organs (96.7%).
• Transfer learning (83.8%) and Convolutional Neural Networks (CNNs, 64.5%) are dominant techniques.
• A significant research gap is the lack of globally available, public datasets for country-specific medicinal plants.

Conclusions:

• Deep learning, particularly CNNs with transfer learning, shows promise for medicinal plant recognition.
• Addressing the need for public, diverse datasets is crucial for advancing the field and ensuring trustworthiness.