Developing pose estimation using a deep learning model to support real-time cow estrus detection

Abstract

Accurate detection of estrus is of paramount importance for the optimization of reproductive efficiency in livestock. Traditional methods are often labor-intensive and subjective. The cow estrus period, which only lasts 12-24 hours in a cycle that repeats every 18-24 days, causes the opportunity to mate or perform artificial insemination to be missed. In this study, we propose a novel approach that utilizes pose estimation with a deep learning model for real-time estrus detection in female cows. We collected a dataset of annotated cow images at different estrus stages and developed a deep learning model based on the EfficientPose architecture. The cow estrus parameter analyzed is locomotion activity which is categorized into several classes (laying, standing, eating, and mounting another cow) with an integrated system and LCD displayed detection result. The input parameter data are processed by the Jetson Nano and YOLO algorithms with an average mean average precision of 0.8, and the lowest loss prediction value of 0.018. If the female cow is classified as active (number of laying class < 57,600 class/hour), then the cow is considered to be in the estrus period. The system provides reliable and noninvasive estrous detection, enabling timely intervention for improved reproductive management in cattle farming.