TCMP-300: A Comprehensive Traditional Chinese Medicinal Plant Dataset for Plant Recognition.

Area of Science:

• Botany
• Computer Science
• Pharmacology

Background:

• Traditional Chinese Medicinal Plants (TCMPs) are vital for disease prevention and treatment.
• Accurate plant recognition is crucial due to diverse therapeutic effects.
• Current expert-dependent identification methods are insufficient for clinical demands, and AI research is hindered by limited datasets.

Purpose of the Study:

• To introduce a comprehensive dataset for Traditional Chinese Medicinal Plants (TCMPs) to advance AI-driven plant recognition.
• To facilitate the development and validation of robust AI models for accurate medicinal plant identification.

Main Methods:

• A dataset of 52,089 TCMP images across 300 categories was curated.
• Images were sourced via Bing search and refined using a pre-trained vision foundation model with human verification.
• State-of-the-art image classification models and advanced data augmentation were employed for technical validation.

Main Results:

• The developed TCMP dataset contains more categories and finer plant part details than existing datasets.
• Technical validation using advanced AI models achieved an accuracy of 89.64% on the dataset.
• The dataset supports comprehensive and fine-grained plant recognition.

Conclusions:

• The presented TCMP dataset significantly enhances AI capabilities for medicinal plant recognition.
• This resource promotes the development and validation of advanced AI models for accurate plant identification in clinical and research settings.