Project-AgML/iNatAg

iNatAg is a large-scale dataset derived from the iNaturalist dataset, designed for species classification and crop/weed classification in agricultural and ecological applications. It consists of 2,959 species with a breakdown of 1,986 crop species and 973 weed species.The dataset contains a total of 4,720,903 images, making it one of the largest and most diverse datasets available for plant species identification and classification.

We also developed a subset of the iNatAg dataset for smaller-scale applications, which we call iNatAg-mini. It contains 560,844 images, created by sampling up to 200 images per species from the full iNatAg dataset. These datasets can then be used in standard agricultural machine learning workflows, enabling the development of extensive applications using this data.

iNatAg and iNatAg-mini are also released as part of the AgML dataset collection, with support for filtering by species, genus, or family and direct data loading through a streamlined API. The associated paper can be found at https://www.arxiv.org/abs/2503.20068.

Installation

To install the latest release of AgML, run the following command:

pip install agml

Quick Start

The following code allows you to list available datasets within the iNatAg dataset:

# List iNatAg species
import agml
print(agml.data.public_data_sources(parent_dataset='iNatAg'))

In line with our vision of providing data across the entire taxonomic hierarchy, we also provide the ability to load data specifically according to family or species, or even common name:

# Load a collection by family names
loader = agml.data.AgMLDataLoader.from_parent("iNatAg", filters={"family_name": ["...", "..."]})

# Load by common names
loader = agml.data.AgMLDataLoader.from_parent("iNatAg", filters={"common_name": "..."})