Pl@ntNet-300K: a new dataset to improve plant images identifications

The initiative aims to enhance research for the improvement of plant images identification at a large scale and based on machine learning. This dataset, built from the database of Pl@ntNet citizen observatory, consists of 306,146 images covering 1,081 species.

The description and evaluation of the dataset have been published in a scientific paper accepted at the Neurips 2021 conference. The dataset itself is publicly available on the Zenodo open research platform.