Multi-Object Tracking in Heterogeneous environments (MOTHe) for animal video recordings

Area of Science:

• Behavioral neuroscience
• Field biology
• Animal behavior research

Background:

• Automated analysis of animal videos is crucial for research.
• Existing tools often fail in natural, heterogeneous environments.
• Field-ready animal tracking methods are frequently inaccessible to researchers.

Purpose of the Study:

• To develop an accessible, open-source tool for animal detection and tracking in challenging natural settings.
• To bridge the gap between advanced computational methods and empirical researchers.

Main Methods:

• Developed MOTHe (Multi-Object Tracking in Heterogeneous environments), a Python package with a graphical interface.
• Utilizes a convolutional neural network for object detection.
• Enables automated training data generation, detection in complex backgrounds, and visual tracking.

Main Results:

• Successfully detected and tracked animals in diverse natural habitats, including wasp colonies and antelope herds.
• Demonstrated effectiveness across six video clips with varying environmental conditions.
• MOTHe runs on basic desktop computers without requiring specialized infrastructure.

Conclusions:

• MOTHe provides an accessible solution for automated animal tracking in challenging field conditions.
• The open-source package empowers researchers to extract valuable data from complex aerial videos.
• Facilitates advancements in animal behavior and field biology studies.