Um novo algoritmo de rastreamento em tempo real para suínos baseado em Deep Learning

Abstract

Ensuring the health and well-being of pigs requires significant effort in terms of labor, material resources, and time. The use of traditional monitoring methods can be stressful for pigs and demand substantial resources from producers, especially in large-scale production systems like industrial pig farming. This practice can have negative impacts on pig health and welfare, as well as the economic profitability of pig production. In this context, the aim of this study was to develop a model using the YOLOv8 architecture to detect and track pigs in a group housing environment. A total of 690 images of pigs housed in groups of nine individuals were used, with the dataset divided into training and validation sets in an 80:20 ratio. With the developed model, it was possible to perform tracking and establish individual identification for each pig. However, the metrics used to evaluate the model's performance yielded unsatisfactory results, highlighting the need to increase the training dataset. Despite these challenges, the model demonstrated good performance in terms of frames per second (FPS), indicating its viability for real-time applications, thus ensuring that the model has potential for practical implementation in pig monitoring and management.