Monitoring activity on an individual level of group-housed pigs using computer vision

Modern welfare definitions not only require that the Five Freedoms are met, but animals should also be able toadapt to changes (i.e. resilience) and reach a state that the animals experience as positive. Resilience is defined asthe ability to cope with or quickly recover from a perturbation. Measuring resilience is challenging since relativelysubtle changes in animal behaviour need to be observed 24/7, which would make human observation impossible.Changes in individual activity showed potential in previous studies to reflect resilience. A computer vision (CV)based tracking algorithm for pigs could potentially measure individual activity, which will be more objective andless time consuming than human observations. The aim of this study was to investigate the potential of stateof-the-art CV algorithms for pig detection and tracking for individual activity monitoring in pigs. Pigs were firstdetected using You Only Look Once v3 (YOLOv3) and were tracked using the Simple Online Real-time Tracking(SORT) algorithm. Two videos, of seven hours each, recorded in a barren and an enriched environment were usedto test the tracking algorithm. Three detection models were proposed using different annotation datasets: a modelwith young pigs where annotated pigs were younger than in the test video, a model with older pigs where annotatedpigs were older than the test video, and a combined model where annotations from younger and older pigs werecombined. The combined detection model performed best with a mean average precision (mAP) of over 99.9%in the enriched environment and 99.7% in the barren environment. Intersection over Union (IOU) exceeded 85%in both environments, indicating a good accuracy of the detection algorithm. The tracking algorithm performedbetter in the enriched environment compared to the barren environment, likely due to the larger space per pig.When false-positive tracks where removed (i.e. tracks not associated with a pig), individual pigs were tracked onaverage for 22.3 minutes in the barren environment and 57.8 minutes in the enriched environment. The averagetrack length varied between 7.1 and 138.3 minutes. Thus, based on tracking-by-detection algorithm using YOLOv3and SORT, individual pigs can be tracked automatically in different environments, but manual corrections may beneeded to keep track of the same individual throughout the video