Overview
Badgers are short-legged omnivores and wild animals, and their existence is in danger in some parts of the world.
The Badger dataset is a small sample subset of videos of different badgers collected by the foundation of
Das and Boom.
The dataset contains four individual instances of badgers with a total number of 51 videos (1641 frames containing the existence of badgers). The badger classes (identities) are: badger_esp, badger_iaco, badger_looi, and badger_strik.
The badgers were recorded in 2016 and 2017 at the Badger Rescue Center of Das & Boom in the Netherlands. Additionally, some videos and photos were made at release locations for badger rehabilitation purposes.
To identify each badger, they are micro-chipped, so the animal can be tracked during captivity and identified after release.
The streaming lengths (Ts) of the videos vary in the range between 15 and 60 s. We extracted approximately a frame per second, for which we developed a script that extracts (Ts ± 2) video frames.
We remark that some frames do not contain the existence of badgers and such frames are not used in our experiments.
A detailed descriptions of the dataset splits can be found in the
paper
Additionally, the entire dataset and meta descriptions of the bounding boxes in the train-test splits can be accessed from this link
(download dataset (subset)).
Publication
Citation
@inproceedings{okafor2018detection,
title={Detection and Recognition of Badgers Using Deep Learning},
author={Okafor, Emmanuel and Berendsen, Gerard and Schomaker, Lambert and Wiering, Marco},
booktitle={International Conference on Artificial Neural Networks},
pages={554--563},
year={2018},
organization={Springer}
}