Replication Data for: Redefining lameness assessment: Constructing lameness hierarchy using crowd-sourced data

Lameness causes pain to dairy cows and economic losses to farmers, but can be difficult to routinely monitor. Despite numerous attempts to develop automatic detection methods, few have been successfully applied on farms. The development of reliable automated methods is likely restricted by the lack of large, labeled training datasets that capture the diversity in lameness cases within and among farms. Additionally, conventional gait scoring methods employed for annotating training videos are subjective and unreliable, adding noise to training data and thus hindering model performance. We propose a novel approach to lameness assessment in dairy cows, leveraging crowd-sourced data to construct a lameness hierarchy. In this pilot study using 30 cow videos, we evaluated the reliability of traditional gait scoring systems revealing their subjective and inconsistent nature, with intra- and interobserver reliabilities of ICC (intraclass correlation coefficient)=0.62 ± 0.09 and 0.44 ± 0.02, respectively. Conversely, our proposed lameness hierarchy constructed from pairwise lameness assessments, achieved a high interobserver reliability (ICC = 0.81) among experienced assessors. We also demonstrated feasibility for the pairwise assessment to be executed by untrained assessors (in this case crowd workers recruited via Amazon MTurk), evidenced by high agreement between hierarchies generated by crowd workers and experienced assessors (ICC = 0.85). Utilizing a 5-milestone subsampling algorithm, we found that recruiting just 8 crowd workers per video pair is sufficient to construct a reliable lameness hierarchy. This method also decreases the number of pairwise comparisons by 61%, relative to evaluating all possible comparisons between every pair of cows. The proposed lameness hierarchy method facilitates quick and accurate labeling of lameness videos and enables a more granular evaluation of lameness. We suggest that this approach can be used to create large training datasets suitable for developing reliable automatic lameness detection models.