islgl/aod-dataset

Animal Object Detection Dataset

AOD Dataset

Introduction

The AOD (Animal Object Detection) dataset is a large-scale animal image dataset specifically designed for object detection tasks. It integrates animal images and bounding box annotations from five publicly available datasets. These datasets encompass both animal-focused datasets and general datasets covering a wide range of scenes.

Our dataset encompasses data from 18 different animal species, including livestock and wildlife. You can find a complete list of categories in categories.yaml. It comprises a total of 35,054 images and 81,078 annotated bounding boxes, capturing scenes ranging from everyday life to zoos and wildlife environments.

The dataset is organized in the YOLO data format, with each image having a corresponding .txt file containing bounding box annotations. For detailed information on YOLO data format, please refer to the Ultralytics YOLOv5 documentation.

Download

You can download the whole dataset from Google Drive, OneDrive or BaiduNetdisk.

Set	Images	Instances
train	28,051	64,834
validation	7,003	16,244

AOD-Rare

In addition, we also provide the AODR (AOD-Rare) dataset, which includes the complete AOD dataset and incorporates an additional 10 rare animal data categories. AODR comprises over 55,000 images and more than 110,000 instances. You can download the AODR dataset from OneDrive.

Acknowledgement

All data in the AOD dataset is sourced from five public datasets, including well-known and widely used datasets such as Pascal VOC 2012¹ and MS COCO 2017². Additionally, the AOD dataset includes data from three animal-focused public datasets, namely Animal Pose³ by Jinkun Cao and others, LoTE-Animal⁴ by Dan Liu and others, as well as ATRW (Amur Tiger Re-identification in the Wild)⁵ from CVWC 2019. We express sincere gratitude to the contributors of these datasets.

Copyright

This dataset is licensed under the CC BY 4.0 license. Refer to the LICENSE file for details.

References

Footnotes

1. Everingham, M., et al, "The PASCAL Visual Object Classes Challenge 2012 (VOC2012) Results," ↩

2. T.-Y. Lin et al., “Microsoft COCO: Common Objects in Context.” arXiv, Feb. 20, 2015. doi: 10.48550/arXiv.1405.0312. ↩

3. Cao, J., et al, "Cross-Domain Adaptation for Animal Pose Estimation," in The IEEE International Conference on Computer Vision (ICCV), 2019. ↩

4. Liu, D., et al, "LoTE-Animal: A Long Time-span Dataset for Endangered Animal Behavior Understanding," in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023, pp. 20064-20075. ↩

5. S. Li, J. Li, H. Tang, R. Qian, and W. Lin, “ATRW: A Benchmark for Amur Tiger Re-identification in the Wild,” in Proceedings of the 28th ACM International Conference on Multimedia, Oct. 2020, pp. 2590–2598. doi: 10.1145/3394171.3413569. ↩