Application of Machine Learning for Automating Behavioral Tracking of Captive Bornean Orangutans (Pongo Pygmaeus).

Area of Science:

• Primate behavior analysis
• Machine learning applications
• Computer vision in zoology

Background:

• Automated behavior tracking is crucial for understanding animal welfare and behavior in captive environments.
• Traditional methods of behavior tracking can be labor-intensive and prone to human error.
• Machine learning offers a promising avenue for developing objective and efficient behavior analysis tools.

Purpose of the Study:

• To investigate the efficacy of object detection using machine learning for automating behavior tracking in captive Bornean orangutans.
• To assess the potential of object detection models in analyzing primate locomotion and filtering false positives.
• To identify areas for improvement in developing robust automated behavior tracking systems.

Main Methods:

• Object detection was applied to CCTV video footage of two captive Bornean orangutans.
• A training dataset of 334 images was extracted from a 2-minute video and labeled using Rectlabel software.
• An object detection model was constructed using the labeled data with Create ML software.

Main Results:

• Object detection demonstrated potential for automating behavior tracking, particularly for locomotion.
• The method showed effectiveness in filtering out false positives, enhancing tracking accuracy.
• The study identified specific areas for model improvement, such as training data diversity and iteration control.

Conclusions:

• Object detection presents a viable tool for automating primate behavior tracking, offering efficiency and objectivity.
• Future implementations should prioritize diverse training datasets and controlled model iterations to prevent overfitting and maximize performance.
• This approach can significantly aid in monitoring and understanding primate behavior in zoological settings.