Meta_Album_MED_LF_Extended: 1,596 Images of 26 Medicinal Leaf Species

1,596 images of healthy leaves from 26 medicinal plant species, captured on a uniform white background. The dataset is a processed version of the Medicinal Leaf Database, with images center-cropped and resized to 128x128 pixels for use in Meta-Album, a benchmark for few-shot image classification. It was created by Phan Anh VU in March 2022, based on original data from S Roopashree and J Anitha released in 2020.

Use Cases

• Benchmarking few-shot image classification models based on the 26 plant species classes.
• Training models for medicinal plant identification from leaf images on a uniform background.
• Developing computer vision systems for automated plant health screening, as the dataset is part of the 'Plant Diseases' domain.
• Studying domain adaptation in meta-learning using a standardized, preprocessed image dataset.

Strengths

• Contains 1,596 images, providing a defined scale for model training and evaluation.
• Images are preprocessed to a uniform 128x128 pixel size and white background, reducing initial data cleaning effort.
• Clear licensing under CC BY 4.0 for both the original and Meta-Album releases.
• Includes 26 distinct classes, offering a multi-class classification challenge.

Limitations

• The specific number of images per class is not provided, which may obscure class balance.
• Original high-resolution detail (1600x1200 px) is lost due to cropping and resizing to 128x128 px.
• Metadata such as column descriptions for potential annotations or sample data is unavailable for inspection.

Provenance

Source

Medicinal Leaf Database on Mendeley Data, authored by S Roopashree and J Anitha.

Collection Method

Leaves were plucked from plants, placed on a white background, and photographed with a mobile phone camera.