Skip to content

biobench.herbarium19

Herbarium19: classify specimens into species based on the 2019 FGVC6 competition.

@article{tan2019herbarium,
  title={The herbarium challenge 2019 dataset},
  author={Tan, Kiat Chuan and Liu, Yulong and Ambrose, Barbara and Tulig, Melissa and Belongie, Serge},
  journal={arXiv preprint arXiv:1906.05372},
  year={2019}
}

Dataset

Bases: ImageFolder

ImageFolder but returns Sample.

Sample

Bases: TypedDict

A dictionary representing a single image sample with its metadata.

Attributes:

Name Type Description
img_id str

Unique identifier for the image.

img Float[Tensor, '3 width height']

The image tensor with shape [3, width, height] (RGB channels first).

label Int[Tensor, '']

Binary class label (0 or 1) for the image.