Predicting severe tail damage in pigs through group and individual activity changes using a computer vision tracking algorithm

Tail biting is a damaging behaviour that can escalate within groups during an outbreak. Early-stage intervention, such as removing the biter or victim, may help to contain the outbreak. While previous research suggests that increased group activity can predict outbreaks, it is unclear whether biters, victims and neutral pigs show distinct activity changes that could guide effective interventions. Detecting these changes requires long-term individual activity monitoring, achievable through tracking algorithms. This study examines whether a computer vision tracking algorithm can predict tail damage in pigs by detecting activity changes before tail damage occurs at both the group and individual level. Nine damage pens were selected based on the criterium that at least one pig had a wound on two consecutive observation days (either D0 and D+3, or D0 and D+4, where D0 refers to the first observation day). Each damage pen was matched with a control pen that showed a low incidence of tail damage on the same days. Tail manipulation, including biter and victim identities, was scored from videos on D-5, D-2, and D+1. A tracking algorithm was applied on several days around the tail damage (D−9, D−5, D−2, and D+1) to calculate the distance moved by each individual. Preliminary results indicate no significant increase in group-level activity in damage pens around tail damage occurrence, nor any difference compared to control pens. However, individual activity at D-2 increases with biting frequency at D-2 and D+1, suggesting that activity may serve as a potential indicator and predictor for identifying biters.