Automated Activity Tracking and Space Use Monitoring of Captive Jaguars with Machine Learning.

Area of Science:

• Zoological research
• Machine learning applications
• Animal behavior analysis

Background:

• Monitoring animal welfare and wildlife conservation is crucial.
• Manual analysis of monitoring data is time-consuming and labor-intensive.
• Machine learning (ML) offers automated solutions to reduce bias and workload.

Purpose of the Study:

• To investigate the activity and space use of captive jaguars using ML.
• To demonstrate the potential of ML for automated individual recognition and activity tracking.
• To assess the feasibility of ML for enhancing animal welfare monitoring in zoological institutions.

Main Methods:

• Trained an ML model on video footage of three captive jaguars (Panthera onca).
• Utilized automated individual recognition, activity tracking, and heatmap visualization.
• Analyzed 123.8 hours of video footage for movement patterns and behavior.

Main Results:

• The ML model achieved satisfactory performance metrics (mean average precision, recall, precision, F1-score).
• Identified repeated movement tracks within specific enclosure areas.
• Observed significantly more inactive than active behavior in jaguars, with no clear nocturnal/crepuscular hunting patterns.

Conclusions:

• ML methods show potential as valuable tools for individual recognition, activity tracking, and space use monitoring.
• This proof-of-concept study highlights ML's utility for future animal welfare assessments.
• Further research with larger sample sizes is warranted to fully validate ML applications in zoological settings.