Meta-Album PlantNet: 120,688 Images of 25 Plant Species

Meta-Album PlantNet is an image dataset containing 120,688 standardized 128x128 pixel images across 25 plant species. The data originates from the Pl@ntNet Project, where citizen botanists worldwide contributed images that were validated by other users. The dataset was created by Felix Herron and released in March 2022 as part of the Meta-Album benchmark collection.

Use Cases

• Training image classifiers to identify plant species based on standardized citizen-science photographs.
• Benchmarking few-shot or meta-learning algorithms on a curated, multi-class plant image dataset.
• Studying the effectiveness of citizen-science data collection and validation for machine learning applications.
• Developing educational tools for plant recognition using a large, labeled image corpus.

Strengths

• Contains 120,688 images, providing substantial volume for model training.
• Includes 25 classes with a balanced distribution, ranging from 2,914 to 9,011 images per class.
• Images are preprocessed to a uniform 128x128 pixel size using a standard anti-aliasing filter.
• Each image has been reviewed by an average of 2.03 citizen botanists, suggesting a validation mechanism.

Limitations

• Description metadata is limited; actual data quality and label accuracy require manual inspection after download.
• The dataset is a subset of a larger collection, which may limit generalizability to a broader set of plant species.
• The image preprocessing (cropping and scaling to 128x128) may have removed contextual details present in the original photographs.

Provenance

Source

Pl@ntNet Project, via the Pl@ntNet-300k dataset.

Collection Method

Images sourced by citizen botanists and validated by other users using a weighted reliability score.

Time Range

null

Freshness

Created on 01 March 2022; last update date is unknown.

Geography

Global, based on contributions from citizen botanists worldwide.