Automated markerless pose estimation in freely moving macaques with OpenMonkeyStudio.

Area of Science:

• Neuroscience
• Psychology
• Ethology
• Medicine
• Primate Research

Background:

• Rhesus macaques are vital models in neuroscience, psychology, ethology, and medicine.
• Precise behavioral measurement in freely moving macaques is crucial for enhancing research utility.
• Current tracking methods lack sufficient precision for unconstrained macaque behavior.

Purpose of the Study:

• To introduce OpenMonkeyStudio, a novel deep learning-based markerless motion capture system.
• To enable precise 3D pose estimation for freely moving macaques in large, unconstrained environments.
• To overcome limitations of existing tracking approaches for macaque behavioral analysis.

Main Methods:

• Utilized 62 machine vision cameras within a 2.45m x 2.45m x 2.75m enclosure.
• Employed multiview image streams for data augmentation via 3D reconstruction.
• Trained a robust, view-invariant deep neural network for automatic pose inference.

Main Results:

• Achieved markerless 3D pose estimation for macaques in unconstrained motion.
• Demonstrated view invariance, a significant improvement over 2D markerless tracking.
• Successfully recognized actions and tracked social interactions with high accuracy.

Conclusions:

• OpenMonkeyStudio provides a powerful tool for detailed behavioral analysis in rhesus macaques.
• The system's view-invariant deep learning approach enables accurate tracking of naturalistic behaviors.
• This technology significantly enhances the utility of the macaque model in scientific research.