BeeDataset

Overview

The BeeDataset contains images of bees annotated with four binary labels. Each label represents a separate attribute, allowing an image to have multiple positive labels.

The dataset provides the following labels:

• varroa

• pollen

• wasps

• cooling

Train: 7,490 images (the dataset has no official validation or test split)

Data Structure

When accessing an example using ds[i], you will receive a dictionary with the following keys:

Key	Type	Description
image	PIL.Image.Image	Image of a bee.
varroa	int	Binary label indicating the presence or absence of varroa.
pollen	int	Binary label indicating the presence or absence of pollen.
wasps	int	Binary label indicating the presence or absence of wasps.
cooling	int	Binary label indicating the cooling-related annotation.

Usage Example

Basic Usage

from stable_datasets.images.bee_dataset import BeeDataset

# Load the training split
ds = BeeDataset(split="train")

sample = ds[0]
print(sample.keys())

# Optional: make it PyTorch-friendly
ds_torch = ds.with_format("torch")

References

• Dataset website: https://github.com/BeeAlarmed/BeeDataset?tab=readme-ov-file

• Tensorflow information source: https://www.tensorflow.org/datasets/catalog/bee_dataset

• Original dataset source - deprecated : https://raspbee.de

• License: GPL-3.0 LICENSE

Citation

@misc{beealarmed2021,
    title = {Dataset for a camera based bee-hive monitoring},
    author = {Fabian Hickert},
    year = {2021},
    journal = {BeeAlarmed},
    url = {https://github.com/BeeAlarmed}
}