Data from neck acceleration tags for training cow's feeding behavior classifier

The dataset includes 3D-acceleration samplings measured by RuuviTag tags. The tags were fitted to 48 cows: 48 to the neck and 48 to a leg. The acceleration was sampled at 25Hz for three axis, and the frequency of the message sending was 5Hz. Each packet included five samples for three axes, which amounted to 15 acceleration values. The data from the tags were received by 6 Raspberry Pi 3 receiving stations, which were single-board computers equipped with a Bluetooth antenna. The receiving stations recorded the tag accelerations and the receiving time and stored the raw data in CSV files (2021-03-04.zip -- 2021-05-15.zip) during 68 days.

The barn was equipped with 24 Hokofarm individual feeders measuring the feeding time and recorded in files in the folder FeedingData in AdditionalData.zip.

Individual cows were recognized by their unique color patterns of coat; images of cows from both sides, on top and from the head were captured at the beginning of the experiment to aid the recognition (Experimental cows.zip).

Three feeding behavior classes were considered in this study: feeding, ruminating and other (neither ruminating nor feeding). The behavior reference included data from feeders (feeding) and manually was generated according to videos (ruminating and other).

The behavior of 21 cows was labeled during three days (4.3.2021, 5.3.2021 and 13.3.2021) and written in csv files Reference_FeedingBehavior_2021-03-...csv in the folder AdditionalData.zip.

The labeled data where acceleration was collected from all receiving stations and fitted to the available references is stored in files AccDataLabeled_Tag.._Cow.._2021-03-...csv in folder Labeled25.zip. These files were used for the training and validation of the classifiers.

Function for reading and processing the data are available on https://github.com/cowbhave/CowBhave_BehaviorModel.

The barn was equipped with continuously recording Dahua 5M cameras installed on the ceiling and covering the major part of the area. The cameras recorded during 4 days. Part of the recorded videos is available on https://zenodo.org/record/6792640 in the video files (Maa9..mp4 -- Maa10..mp4).