Intelligent Poultry Health Management Based on Improved YOLOv8 Algorithm

Poultry production plays a critical role in the development of the global economy. Meanwhile, it also raises food safety concerns for humans. Poultry disease prevention and control are important for farmers to ensure food safety and economic profits. In this study, an abnormal poultry behavior detection network was proposed and developed based on an improved YOLOv8 deep learning algorithm, leveraging the advantages of modern technologies to enhance the efficiency of non-invasive monitoring for poultry health management. The YOLOv8 model was improved by combining the Alterable Kernel Convolution (AKConv) computational block and the Efficient Multiscale Attention (EMA) mechanism block. Based on experimental results, a 99.2% detection precision of the proposed improved YOLOv8 model was achieved, significantly better than other typical deep learning algorithms, such as Retinanet, YOLOv3, YOLOv5, YOLOv8, and YOLOv9. This provides effective technical support for poultry health management by detecting abnormal behaviors in poultry.