Behavioral observation is central to animal welfare assessment in zoological institutions, but manual observation is time-consuming and difficult to apply continuously. We developed and applied a LabGym-based workflow for automated behavioral monitoring of captive lions (Panthera leo). Development footage from three Danish zoological institutions was used to train a lion detector and a seven-class behavior categorizer; video collected on 22–28 December 2025 was temporally held out from model development and used for applied validation. Detector performance was moderate (bounding-box AP 47.1%; segmentation AP 41.6%), while the internal categorizer validation reached 0.62 accuracy and weighted F1. In the held-out validation, 52.6% of manually annotated lion-time received a retained pipeline prediction. Fine-scale end-to-end accuracy was 0.231 and increased to 0.440 when calculated only on covered lion-time. Grouping behaviors as active, inactive/maintenance, and pacing-like locomotion increased these values to 0.317 and 0.602, respectively. Coverage varied substantially among institutions (33.8–59.8%) and was particularly low for cubs and nighttime footage. Although model macro-F1 exceeded simple majority-class baselines, end-to-end accuracy did not. The workflow demonstrates the potential of computer-vision approaches for scalable behavioral quantification and future welfare assessment in zoological settings, while highlighting the importance of accounting for prediction coverage and classification performance when interpreting automated behavioral data.

Behavioral monitoring is central to welfare assessment but is difficult to conduct continuously. We used machine-learning-based video analysis to describe lion (Panthera leo) behavior across three zoo enclosures differing in available space, structural complexity and social composition. Interpretation of the Aalborg Zoo profile is additionally constrained by a major recent social change, as the enclosure’s adult male was euthanized five days before data collection began. Continuous recordings collected over 10 days were processed with LabGym, filtered for low-confidence predictions and likely static false detections, and summarized as behavioral durations. Because enclosure characteristics, pride size and cub presence varied only between three enclosures, enclosure was treated as the independent comparative unit; repeated hourly and daily observations were used to characterize within-enclosure temporal variation rather than to generate inferential replication. Analyses therefore focused on time budgets, daily and pooled activity, Shannon and Simpson diversity, within- and between-day behavioral instability, social interaction and pacing-like movement. The three enclosures showed distinct behavioral profiles. Diversity was lowest in Aalborg Zoo, social interaction did not increase consistently with pride size, and pacing-like movement was concentrated in Aalborg Zoo. However, the pacing-like measure should be interpreted as a screening indicator, as only 27.7% of manually annotated pacing-like lion-time was retained as pacing-like by the complete pipeline; it therefore cannot be interpreted as an estimate of true pacing prevalence. Activity was highest in Aalborg Zoo, but did not vary consistently with available space or structural complexity across the three enclosures. Automated quantification can support high-resolution exploratory welfare monitoring, but broader inference requires substantially more independently replicated enclosures.