Sheep videos taken from drone at low altitude

This dataset was developed within the framework of the European Horizon 2020 project ICAERUS, specifically for the livestock monitoring use case. The objective of this work is to explore the potential of drone-based computer vision methods for monitoring small ruminants in real farming environments. More information about the project is available on the project website: https://icaerus.eu Objective Counting sheep and goats is a significant operational challenge for farmers managing flocks that may contain hundreds of animals. Traditional counting methods are time-consuming and prone to errors. The objective of this work is to develop a computer vision–based methodology capable of automatically detecting, tracking, and counting sheep and goats when animals pass through a corridor, gate, or other naturally constrained passage. The proposed approach relies on low-altitude aerial videos (<15 m) acquired using drones, providing a top-down perspective that facilitates the detection and counting of animals. Progress and Enhancements Our work includes the development of datasets and models dedicated to low-altitude aerial imagery of sheep (<15 m). Datasets contributions: Multiple datasets either with or without annotations, have been produced and enriched as part of this work during the 2023-2026 period (see the summary table). Name Version Date Link How to quote ? Number of Images Number of Videos Number of Bounding Boxes Drone raw images of cattle in french grazing areas v1 10-08-2023 https://zenodo.org/records/8234156 Lebreton, A. (2023). Drone raw images of cattle in french grazing areas [Data set]. Zenodo. https://doi.org/10.5281/zenodo.8234156 900 Drone images and their annotations of grazing cows v1 01-12-2023 https://zenodo.org/records/10245396 Lebreton, A., & Helary, L. (2023). Drone images and their annotations of grazing cows [Data set]. Zenodo. https://doi.org/10.5281/zenodo.10245396 1100 Drone images and their annotations of grazing cows v2 01-04-2024 https://zenodo.org/records/11048412 Helary, L., & Lebreton, A. (2024). Drone images and their annotations of grazing cows [Data set]. Zenodo. https://doi.org/10.5281/zenodo.11048412 1385 4941 Sheep videos taken from drone at low altitude v1 18-12-2023 https://zenodo.org/records/10400302 Lebreton, A., & Helary, L. (2023). Sheep videos taken from drone at low altitude [Data set]. Zenodo. https://doi.org/10.5281/zenodo.10400302 16 Drone videos and their annotations of passing sheep (for counting purpose) v1 18-06-2024 https://zenodo.org/records/12094356 Helary, L., Okoye, K. N., Kolodziejczyk, M., Schewe, J., Philip, L., Nicolas, E., & Lebreton, A. (2024). Drone videos and their annotations of passing sheep (for counting purpose) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.12094356 4 14365 Aerial videos and images of goats (for computer vision purpose) v1 03-01-2025 https://zenodo.org/records/14591324 Lebreton, A., Depuille, L., Nicolas, E., & Helary, L. (2025). Aerial videos and images of goats (for computer vision purpose) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.14591324 2056 10 Drone images and their annotations of goats/small ruminants (for computer vision purpose) v1 26-02-2025 https://zenodo.org/records/14929694 Lebreton, A., Duval, L., Depuille, L., Nicolas, E., & Helary, L. (2025). Drone images and their annotations of goats/small ruminants (for computer vision purpose) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.14929694 287 2790 Drone videos and images of sheep in various conditions (for computer vision purpose) v1 04-03-2025 https://zenodo.org/records/14967219 Lebreton, A., Morin, C., Nicolas, E., & Helary, L. (2025). Drone videos and images of sheep in various conditions (for computer vision purpose) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.14967219 1315 28 Drone videos and images of sheep in various conditions (for computer vision purpose) - Part II v1 06-03-2026 https://zenodo.org/records/18889354 Lebreton, A., Helary, L., NICOLAS, E., Goin, L., Grisot, P.-G., & Jegorel, T. (2026). Drone videos and images of sheep in various conditions (for computer vision purpose) - Part II [Data set]. Zenodo. https://doi.org/10.5281/zenodo.18889354 1679 47 Drone images and their annotations of sheep in various conditions (for computer vision purpose) v1 06-03-2026 https://zenodo.org/records/18889623 Lebreton, A., de Brito, A., Blaise, E., Jegorel, T., Goin, L., Grisot, P.-G., NICOLAS, E., & Helary, L. (2026). Drone images and their annotations of sheep in various conditions (for computer vision purpose) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.18889623 809 18018 Drone videos to test sheep counting computer vision counting pipeline v1 06-03-2026 https://zenodo.org/records/18889878 Lebreton, A., Grisot, P.-G., Depuille, L., Goin, L., NICOLAS, E., & Helary, L. (2026). Drone videos to test sheep counting computer vision pipeline [Data set]. Zenodo. https://doi.org/10.5281/zenodo.18889878 98 TOTAL 9531 203 40114 Model Development: We developed computer vision models for small ruminant detection (0.99 mAP50 in its version 4), tracking, and counting.The models and associated code are available on GitHub:https://github.com/ICAERUS-EU/UC3_Livestock_Monitoring To improve the performance and robustness of detection models such as YOLO, the datasets were enriched to increase variability in: Environmental conditions (background types and lighting conditions) Animal appearance, including non-white sheep and goats, which are often underrepresented in existing datasets. Data set description This first dataset will support our work. The dataset encompasses 16 .MP4 videos from drone (DJI mavic 3 Enterprise or Thermal) of around 50 sheep crossing a gate. The videos were taken from 5m to 10m of height and to an horizontal distance of the gate from 0m to 10m. Future Work Following extensive efforts in data collection and annotation, our next objective is to finalize and deploy the sheep counting pipeline on an edge computing solution, enabling real-time livestock monitoring in operational farm environments. In parallel, additional projects are exploring other computer vision applications in sheep farming, expanding the potential use cases of this technology. Acknowledgments The authors also thank all the farm staff and technical teams involved in the data acquisition campaigns for their assistance in enabling drone flights and data collection under real farming conditions. Collaboration and Contact We welcome collaborations on this topic. For inquiries or further information, please contact:Adrien LebretonEmail: [e-mail]本数据集是在欧盟地平线2020（European Horizon 2020）计划ICAERUS项目框架下开发的，专门面向畜牧监测应用场景。本研究的目标是探索基于无人机（drone）的计算机视觉（computer vision）方法在真实农业环境中监测小型反刍动物的潜力。 有关该项目的更多信息可访问项目官网：https://icaerus.eu ## 研究目标 对于饲养数百只牲畜的养殖户而言，清点绵羊和山羊的数量是一项极具挑战性的日常工作。传统的计数方法不仅耗时耗力，还极易出现差错。 本研究旨在开发一种基于计算机视觉的方法，能够在牲畜通过通道、围栏门或其他天然受限通道时，自动检测、追踪并计数绵羊和山羊。 本研究所提出的方案依托无人机采集的低空航拍视频（高度低于15米），通过自上而下的拍摄视角，简化了牲畜的检测与计数流程。 ## 进展与优化 本研究的工作内容涵盖针对绵羊低空航拍（高度低于15米）的数据集与模型开发。 ### 数据集贡献 本研究在2023-2026年期间生成并丰富了多组带标注或无标注的数据集（详见汇总表格）。 | 名称 | 版本 | 日期 | 链接 | 引用方式 | 图像数量 | 视频数量 | 边界框数量 | | --- | --- | --- | --- | --- | --- | --- | --- | | 法国牧场区域牛群无人机原始影像 | v1 | 2023年8月10日 | https://zenodo.org/records/8234156 | Lebreton, A. (2023). 法国牧场区域牛群无人机原始影像 [数据集]. Zenodo. https://doi.org/10.5281/zenodo.8234156 | 900 | 无 | 无 | | 放牧奶牛无人机影像及标注数据 | v1 | 2023年12月1日 | https://zenodo.org/records/10245396 | Lebreton, A., & Helary, L. (2023). 放牧奶牛无人机影像及标注数据 [数据集]. Zenodo. https://doi.org/10.5281/zenodo.10245396 | 1100 | 无 | 无 | | 放牧奶牛无人机影像及标注数据 | v2 | 2024年4月1日 | https://zenodo.org/records/11048412 | Helary, L., & Lebreton, A. (2024). 放牧奶牛无人机影像及标注数据 [数据集]. Zenodo. https://doi.org/10.5281/zenodo.11048412 | 1385 | 无 | 4941 | | 低空无人机拍摄的绵羊视频 | v1 | 2023年12月18日 | https://zenodo.org/records/10400302 | Lebreton, A., & Helary, L. (2023). 低空无人机拍摄的绵羊视频 [数据集]. Zenodo. https://doi.org/10.5281/zenodo.10400302 | 无 | 16 | 无 | | 用于计数的过路绵羊无人机视频及标注数据 | v1 | 2024年6月18日 | https://zenodo.org/records/12094356 | Helary, L., Okoye, K. N., Kolodziejczyk, M., Schewe, J., Philip, L., Nicolas, E., & Lebreton, A. (2024). 用于计数的过路绵羊无人机视频及标注数据 [数据集]. Zenodo. https://doi.org/10.5281/zenodo.12094356 | 无 | 4 | 14365 | | 用于计算机视觉研究的山羊航拍视频与图像 | v1 | 2025年1月3日 | https://zenodo.org/records/14591324 | Lebreton, A., Depuille, L., Nicolas, E., & Helary, L. (2025). 用于计算机视觉研究的山羊航拍视频与图像 [数据集]. Zenodo. https://doi.org/10.5281/zenodo.14591324 | 2056 | 10 | 无 | | 用于计算机视觉研究的山羊/小型反刍动物无人机影像及标注数据 | v1 | 2025年2月26日 | https://zenodo.org/records/14929694 | Lebreton, A., Duval, L., Depuille, L., Nicolas, E., & Helary, L. (2025). 用于计算机视觉研究的山羊/小型反刍动物无人机影像及标注数据 [数据集]. Zenodo. https://doi.org/10.5281/zenodo.14929694 | 287 | 无 | 2790 | | 用于计算机视觉研究的多场景绵羊无人机视频与图像 | v1 | 2025年3月4日 | https://zenodo.org/records/14967219 | Lebreton, A., Morin, C., Nicolas, E., & Helary, L. (2025). 用于计算机视觉研究的多场景绵羊无人机视频与图像 [数据集]. Zenodo. https://doi.org/10.5281/zenodo.14967219 | 1315 | 28 | 无 | | 用于计算机视觉研究的多场景绵羊无人机视频与图像——第二部分 | v1 | 2026年3月6日 | https://zenodo.org/records/18889354 | Lebreton, A., Helary, L., NICOLAS, E., Goin, L., Grisot, P.-G., & Jegorel, T. (2026). 用于计算机视觉研究的多场景绵羊无人机视频与图像——第二部分 [数据集]. Zenodo. https://doi.org/10.5281/zenodo.18889354 | 1679 | 47 | 无 | | 用于计算机视觉研究的多场景绵羊无人机影像及标注数据 | v1 | 2026年3月6日 | https://zenodo.org/records/18889623 | Lebreton, A., de Brito, A., Blaise, E., Jegorel, T., Goin, L., Grisot, P.-G., NICOLAS, E., & Helary, L. (2026). 用于计算机视觉研究的多场景绵羊无人机影像及标注数据 [数据集]. Zenodo. https://doi.org/10.5281/zenodo.18889623 | 809 | 无 | 18018 | | 用于测试绵羊计数计算机视觉流程的无人机视频 | v1 | 2026年3月6日 | https://zenodo.org/records/18889878 | Lebreton, A., Grisot, P.-G., Depuille, L., Goin, L., NICOLAS, E., & Helary, L. (2026). 用于测试绵羊计数计算机视觉流程的无人机视频 [数据集]. Zenodo. https://doi.org/10.5281/zenodo.18889878 | 无 | 98 | 无 | | 总计 | - | - | - | - | 9531 | 203 | 40114 | ## 模型开发 我们开发了用于小型反刍动物检测、追踪与计数的计算机视觉模型（第4版本的mAP50达到0.99）。相关模型与代码已开源至GitHub：https://github.com/ICAERUS-EU/UC3_Livestock_Monitoring 为提升YOLO等检测模型的性能与鲁棒性，本研究对数据集进行了扩充，以增加以下维度的多样性： 1. 环境条件（背景类型与光照条件） 2. 牲畜外观特征，包括现有数据集中常被遗漏的非白色绵羊与山羊。 ## 数据集说明 本首份数据集将为我们的研究提供支撑。 该数据集包含16段.MP4格式的无人机航拍视频，拍摄设备为DJI Mavic 3 Enterprise或DJI Mavic 3 Thermal，内容为约50只绵羊穿越围栏门的场景。视频拍摄高度介于5米至10米之间，与围栏门的水平距离介于0米至10米之间。 ## 未来工作 在完成大量的数据采集与标注工作后，我们的下一目标是完成绵羊计数流程的开发，并将其部署至边缘计算方案中，从而在实际农场环境中实现实时畜牧监测。 与此同时，其他相关项目正在探索计算机视觉在绵羊养殖中的其他应用场景，进一步拓展该技术的潜在应用范围。 ## 致谢 作者谨向所有参与数据采集工作的农场工作人员与技术团队致谢，感谢他们协助在真实农业环境中开展无人机飞行与数据采集工作。 ## 合作与联系方式 我们欢迎就该主题开展合作。如有咨询或获取更多信息的需求，请联系：Adrien Lebreton 邮箱：[e-mail]