There is a newer version of the record available.

Published December 1, 2023 | Version 1
Dataset Open

Drone images and their annotations of grazing cows

  • 1. Institut de l'Elevage
  • 2. ROR icon Institut de l’Elevage

Description

The dataset is part of the European H2020 project ICAERUS regarding the livestock monitoring use case. More information here : https://icaerus.eu/

This dataset was built to develop cow detection and counting models using drone images. 

The directory contains the dataset encompasses around 1100 raw .jpeg images from drone (DJI mavic 3 Enterprise or Thermal) of grazing areas where cattle graze collected between June and October 2023.

The nadir images were collected during flight planned with DJI Pilot 2 at a constant altitude regarding the take-off position.  

Images were taken from altitudes of 30m, 60m and mostly 100m. The Aerial images contain many .exif methadata (drone, camera, altitude etc.).

There is a strong imbalance between images with "cattle" and image with "no cattle" representative of areas to monitor and weather conditions.  

 

Data organization :

The data are organized as first in a directory by farm (Mauron, Derval or Jalogny) where the images were collected and then by a second directory called JPGImages and finally another directory by  

flight planned. Images can be found there.

Name of the directory of each planned flight is defined such as DJI_YYYYMMDDHHMM_XXX with the date (YYYYMMDD), the hour in UTC+2 (HHMM), and XXX representing a mission number.  

An annotation directory is also present in flight directory and contains label files. The name of the annotation directory depends on the type of the annotation format used  

(for now : PASCAL_VOC1.1  and YOLO_V1). One annotation file referred to an unique image and have the same name except the extension.  

Annotations files contain bounding boxes coordinates produced using CVAT.io. Depending on the using model it can be preferable to use PASCAL_VOC1.1 or YOLO_V1 annotations.

 

Other versions of this dataset will be available in 2024. The authors of the dataset are opened to any collaboration regarding animal counting models.

For more information, please contact: adrien.lebreton@idele.fr 

Files

Cattle_drone_images.zip

Files (13.2 GB)

Name Size Download all
md5:b2dcb59e3cc46022372e2a6624e7e7f5
13.2 GB Preview Download

Additional details

Funding

ICAERUS – Innovations and Capacity building in Agricultural Environmental and Rural Uav Services 101060643
European Commission