Leveraging Thermal Infrared Imaging for Pig Ear Detection Research: The TIRPigEar Dataset and Performances of Deep...

Area of Science:

• Agricultural Science
• Computer Vision
• Animal Science

Background:

• Pig ears possess unique thermal properties due to their vascularity, reflecting temperature variations.
• Thermal infrared imaging offers a non-invasive method for assessing individual pig status, despite ears not reflecting core body temperature directly.
• Existing datasets lack comprehensive thermal infrared imagery for pig ear analysis.

Purpose of the Study:

• To introduce the TIRPigEar dataset, a large-scale collection of thermal infrared images of pig ears.
• To facilitate the development and validation of AI models for pig ear detection and temperature analysis.
• To support advancements in precision livestock farming through accessible data.

Main Methods:

• A dataset of 23,189 thermal infrared images (TIRPigEar) was created using an automated inspection robot.
• Images were manually annotated, generating 69,567 labeled files for pig ear detection model training.
• The dataset's efficacy was evaluated using various object detection algorithms, including YOLOv9m.

Main Results:

• The TIRPigEar dataset achieved high performance metrics when integrated with the YOLOv9m object detection model.
• Precision, recall, and mAP50 scores reached 97.35%, 98.1%, and 98.6% respectively with YOLOv9m.
• The dataset proved effective for training AI models, demonstrating the viability of thermal imaging for pig ear analysis.

Conclusions:

• The TIRPigEar dataset is a valuable resource for AI and precision livestock farming research.
• Thermal infrared imaging combined with AI offers a rapid, non-contact, and effective approach for pig health monitoring.
• This dataset will accelerate research in precision livestock farming by enabling efficient pig ear temperature analysis.