Accelerometers and annotated video recordings of behaviours: a dataset for training behaviour classification models in goats

Eight dairy goats were equipped with accelerometers attached to the RFID ear tag of the animal. The accelerometers were programmed to record the acceleration on the x-y-z-axis at a frequency of 5 Hz for 24 consecutive hours. “Ruminating”, “head in the feeder”, “lying”, and “standing” behaviours were annotated from video recordings using the Observer XT software for 11 hours by a single trained observer using a pre-established ethogram. Raw acceleration data and associated behaviours (accessible online) are referred to as the input data for training the algorithm. The ACT4Behav (Accelerometer-based Classification tool for identifying Behaviours) pipeline was used to pre-process this raw data to train behaviour classification models (accessible online). Four models (accessible online) have been pre-trained for each of the four behaviours using the ACT4Behav pipeline.