Precision Livestock Farming Advances with Dairy DigiD

Researchers at Dalhousie University have developed Dairy DigiD, a deep learning-based biometric classification framework that categorizes dairy cattle into four physiologically defined groups using high-resolution facial images. The system combines two approaches: DenseNet121 for full-image classification and Detectron2 for fine-grained facial analysis. Dairy DigiD has achieved classification accuracies between 93 and 98%, demonstrating its superiority over traditional identification methods.

Key Takeaways:

• Dairy DigiD uses a keypoint-driven strategy to refine facial localization and improve classification robustness, enabling robust feature localization and resilience to occlusions, lighting variations, and heterogeneous backgrounds.
• The system integrates animal-centric design with scalable AI, offering an ethical, accurate, and practical alternative to traditional identification methods in precision dairy management.
• Dairy DigiD sets a precedent for responsible, data-driven decision-making in precision dairy management, combining animal-centric design with scalable AI.
• The research concluded that Dairy DigiD represents a significant advancement in automated livestock monitoring, enabling improved animal welfare and more accurate data under real-world farm conditions.
• The system can categorize dairy cattle into four physiologically defined groups: young, mature milking, pregnant, and dry cows.
• Dairy DigiD uses high-resolution facial images, combining two complementary approaches: DenseNet121 for full-image classification and Detectron2 for fine-grained facial analysis.
• The system was tested under uncontrolled farm environments, demonstrating its adaptability and robustness in real-world conditions.

Statistics:

• Dairy DigiD achieved classification accuracies between 93 and 98%.
• The system uses 30 anatomical landmarks (eyes, ears, muzzle) to refine facial localization and improve classification robustness.
• DenseNet121 delivered strong baseline performance, but its sensitivity to background noise limited generalizability.
• Detectron2, on the other hand, demonstrated superior adaptability and robustness in uncontrolled farm environments.

Sources:

• Dairy DigiD: a keypoint-based deep learning system for classifying dairy cattle by physiological and reproductive status. Frontiers in Artificial Intelligence, 2025, 8.
• https://doi-org.sdpl.idm.oclc.org/10.3389/frai.2025.1545247
• NewsRx. Dalhousie University Researchers Update Current Study Findings on Agriculture (Dairy DigiD: a keypoint-based deep learning system for classifying dairy cattle by physiological and reproductive status). Agriculture Week. September 11, 2025; p 22.