Ethoflow: Computer Vision and Artificial Intelligence-Based Software for Automatic Behavior Analysis

Area of Science:

• Ethology
• Computer Science
• Artificial Intelligence

Background:

• Manual animal behavior monitoring is labor-intensive and subjective.
• Existing automated systems struggle with heterogeneous environments and group tracking.
• There's a need for robust software for field-based, multi-individual behavioral analysis.

Purpose of the Study:

• To develop Ethoflow, an AI-powered software for automated animal behavior monitoring.
• To enable accurate tracking and complex behavior assessment in challenging field conditions.
• To provide a flexible tool for biological and ethological research.

Main Methods:

• Implemented an object detection algorithm using instance segmentation for robust tracking.
• Utilized a convolutional neural network (CNN) for complex behavior analysis.
• Developed heuristics for automatic AI model training data generation.

Main Results:

• Ethoflow achieved high accuracy in detecting individuals in heterogeneous environments.
• The software successfully assessed complex behaviors and performed kinematic assessments.
• Demonstrated efficacy in detecting trophallaxis in social bees.

Conclusions:

• Ethoflow offers an automated, accurate solution for animal behavior analysis in diverse settings.
• The software's modular design allows for efficient processing on standard and high-performance hardware.
• Ethoflow is a valuable tool for advancing research in biology and related fields.