FBD-SV-2024

The Flying Bird object Dataset in Surveillance Video (FBD-SV-2024) for surveillance video flying bird object detection algorithm development and performance evaluation.
The FBD-SV-2024 comprises 483 video clips, totaling 28,694 frames of images. Among these frames, 23,833 frames contain a total of 28,366 instances of flying birds.
Visit the GitHub repository for more data processing scripts: https://github.com/Ziwei89/FBD-SV-2024_github

Training set: 400 video clips, totaling 23,979 frames of images.
Testing set: 83 video clips, totaling 4,715 frames of images.

Due to the large size of image data, no images have been uploaded, but relevant Python scripts are provided. These scripts can be run to extract frames from the videos. See the readme_English.txt for more information.

Please refer to the paper related to this dataset for more description of the characteristics of this dataset. The link of the paper : https://arxiv.org/abs/2409.00317
Please refer to the paper if the dataset is useful for your work:
@misc{sun2024fbdsv2024,
title={FBD-SV-2024: Flying Bird Object Detection Dataset in Surveillance Video},
author={Zi-Wei Sun and Ze-Xi Hua and Heng-Chao Li and Zhi-Peng Qi and Xiang Li and Yan Li and Jin-Chi Zhang},
year={2024},
eprint={2409.00317},
archivePrefix={arXiv},
primaryClass={cs.CV}
}

Note: This dataset is for learning and research purposes only, not for commercial use.