Citation

CUB-200-2011 dataset

EfficientNet-B0 and ResNet-50 CUB models use CUB-200-2011. BirdBrain Voyager uses the iNaturalist 2021 birds subset.

The Caltech-UCSD Birds-200-2011 Dataset
Wah, C.; Branson, S.; Welinder, P.; Perona, P.; Belongie, S.
California Institute of Technology, 2011.
Technical Report CNS-TR-2011-001

Dataset home: https://www.vision.caltech.edu/datasets/cub_200_2011/

CUB images are restricted to non-commercial research and educational use. See the official dataset website for terms.

iNaturalist 2021 dataset

iNat training experiments use the FGVC8 challenge data; Birdbrain filters to birds (Aves) only.

Grant Van Horn and macaodha. iNat Challenge 2021 - FGVC8.
https://kaggle.com/competitions/inaturalist-2021, 2021. Kaggle.

Follow the Kaggle competition and iNaturalist dataset terms. This project uses the data for research and educational fine-tuning only—not for commercial products or redistribution of the raw images.

BibTeX (CUB)

@techreport{WelinderEtal2010,
  Author = {C. Wah and S. Branson and P. Welinder and P. Perona and S. Belongie},
  Title = {Caltech-UCSD Birds-200-2011},
  Institution = {California Institute of Technology},
  Year = {2011},
  Number = {CNS-TR-2011-001}
}

BibTeX (iNat 2021)

@misc{inaturalist2021,
  author = {Grant Van Horn and macaodha},
  title = {iNat Challenge 2021 - {FGVC8}},
  howpublished = {\url{https://kaggle.com/competitions/inaturalist-2021}},
  year = {2021},
  publisher = {Kaggle}
}

Model weights

Checkpoints use ImageNet-pretrained backbones from torchvision (EfficientNet-B0, ResNet50). Cite torchvision / the original backbone papers if you publish results that depend on those weights.