Analysis and Comparison of New-Born Calf Standing and Lying Time Based on Deep Learning

Area of Science:

• Animal behavior science
• Veterinary medicine
• Computer vision applications in agriculture

Background:

• Standing and lying durations are key indicators of calf health and welfare.
• Traditional monitoring methods for calf behavior can be labor-intensive and invasive.

Purpose of the Study:

• To develop and validate a computer vision system for non-invasively monitoring calf standing and lying behaviors.
• To analyze the relationship between calf behavior patterns and health status, specifically identifying signs of diarrhea.

Main Methods:

• Utilized a computer vision approach with cameras from four viewpoints to monitor six calves over six days.
• Trained the YOLOv8n model to detect standing and lying calf postures.
• Analyzed daily behavioral budgets using automatic inference on untrained data.

Main Results:

• Achieved a mean average precision of 0.995 for calf behavior detection with an inference speed of 333 frames per second.
• Demonstrated a maximum error of less than 14 minutes in estimated daily standing and lying times across 8 calf-days.
• Identified that calves with diarrhea exhibited significantly longer lying times (approx. 2 hours more) and more frequent lying bouts compared to healthy calves.

Conclusions:

• The proposed computer vision method provides accurate and efficient non-invasive monitoring of calf behavior.
• Automated measurement of standing and lying time can serve as a valuable tool for assessing calf health status.
• This technology has the potential to improve calf welfare and farm management practices.