Tag 'n' Track: Tackling the validation challenge in animal behaviour studies through automated referencing with ArUco markers

Technological advances promise to greatly assist the study of animal behaviour, but the validation of these technologies is often neglected due to its tedious and labour-intensive nature. This paper addresses the challenges of manual annotation in validating technological tools for animal behaviour research. We detail the implementation and effectiveness of a computer vision method that automatically annotates animals within various regions of interest (ROIs). This method uses ArUco markers, open-source visual markers with a grid pattern, fitted onto vests worn by the animals. To validate this method, we used 245 10-minute videos capturing animals’ visits to key resources, using a mobile barn housing twenty-one chickens. Our method generates annotated videos, revealing unique IDs of individuals and timestamps marking their presence in ROIs. Compared with traditional human observation, our method performed excellently: Spearman's correlation (ρ = 0.96, p < 0.01), 92.83 % sensitivity, 99.93 % specificity, 99.08 % accuracy, 98.77 % precision, and a 95.28 % F1-score. All recordings were annotated automatically in 40.96 h, saving 82.72 % of the time compared to the 222.84 h required for manual annotation. The proposed ArUco marker-based tracking method is easy to set up, based on open-source technology, and accessible to researchers without advanced programming skills. This method has the potential to replace or complement manual annotation, simplifying the validation of new technologies for automated individual tracking.