Biosafety Management System for Pig Farms Based on Deep Learning and Internet of Things Technology

African swine fever has had a significant impact on China's pig farms. At present, there is no safe and effective vaccine for African swine fever, and the most effective way to combat it is to establish stringent and safe biological prevention and control measures. This paper proposes a regulatory system based on deep learning and the Internet of Things to minimize the damage caused by African swine fever by supervising personnel, vehicles, and pigs on farms and prompting farm staff to adopt a strict elimination philosophy. In this paper, to address the lack of site-specific datasets for large-scale farms, a large amount of data was collected using cameras in pig farms and selfconstructed farm environment datasets. To deploy the object detection network on the Jetson Nano, the YOLOv5s network is improved and trained in this paper to obtain the optimal weighting parameters. It was found that the improved YOLOv5s network achieved 89.6%, 96.4%, and 92.6% detection accuracy for detecting people, vehicles, and pigs respectively in the pig farm environment, which is 3.2%, 1.6%, and 2.7% higher than the recognition accuracy of the original YOLOv5s network respectively. In addition, this paper implements an intelligent biosafety management system in pig farms by building an Internet of Things database and constructing front-end web pages to receive and display abnormal information reported by the Jetson Nano.