Pl@ntNet-300K: a plant image dataset with high label ambiguity and a long-tailed distribution
Garcin, Camille and Joly, Alexis and Bonnet, Pierre and Affouard, Antoine and Lombardo, Jean-Christophe and Chouet, Mathias and Servajean, Maximilien and Lorieul, Titouan and Salmon, Joseph

@inproceedings{garcin2021plantnet300k,
title= {Pl@ntNet-300K: a plant image dataset with high label ambiguity and a long-tailed distribution},
author= {Garcin, Camille and Joly, Alexis and Bonnet, Pierre and Affouard, Antoine and Lombardo, Jean-Christophe and Chouet, Mathias and Servajean, Maximilien and Lorieul, Titouan and Salmon, Joseph},
booktitle= {Thirty-fifth Conference on Neural Information Processing Systems, Datasets and Benchmarks Track},
year= {2021},
url= {https://zenodo.org/records/5645731},
doi= {10.5281/zenodo.5645731},
keywords= {Classification, images, zenodo, plants, species identification},
license= {https://creativecommons.org/licenses/by/4.0/},
abstract= {**Pl@ntNet-300K** is an image dataset of **306,146 plant images covering 1,081 species**, built from the Pl@ntNet citizen observatory. It has high intrinsic label ambiguity and a long-tailed class distribution.

## What is in this torrent

A single file, `plantnet_300K.zip` (31.7 GB), with the images.

## Why it is interesting

- Strong class imbalance: a few species account for most of the images.
- Many species look alike, so identification is hard even for experts.

These properties make it well suited for evaluating set-valued classification. The authors recommend two metrics, macro-average top-k accuracy and macro-average average-k accuracy, and give baseline results from deep networks trained with cross-entropy.

## Links

- Paper: https://openreview.net/forum?id=eLYinD0TtIt
- Code to load the data and train PyTorch models: https://github.com/plantnet/PlantNet-300K/
- Original record: https://zenodo.org/records/5645731
- DOI: 10.5281/zenodo.5645731

## License

CC BY 4.0. Attribution is required.

## Citation

Please cite:

    Garcin, C., Joly, A., Bonnet, P., Affouard, A., Lombardo, J.-C., Chouet, M., Servajean, M., Lorieul, T., Salmon, J.
    Pl@ntNet-300K: a plant image dataset with high label ambiguity and a long-tailed distribution.
    NeurIPS 2021, Datasets and Benchmarks Track.
  },
terms= {},
superseded= {}
}