Lightweight Sheep Head Detection and Dynamic Counting Method Based on Neural Network

SIMPLE SUMMARY: In the realm of intelligent animal husbandry, the fundamental underpinning of intelligence lies in the capability to identify individual livestock and perform automatic headcounts. Presently, many farming operations still rely on manual counting methodologies, which exhibit notable deficiencies, particularly when confronted with the challenges posed by the substantial populations of sheep and the frequent need for counting. Manual counting is marred by inefficiency and susceptibility to errors such as duplication and omission, thereby presenting formidable hurdles. In response to these challenges within the domain of animal husbandry, this research introduces a deep neural network model that leverages contactless computer vision technology for the automatic detection and enumeration of sheep. We have systematically deployed a series of advanced and efficacious strategies to enhance the model’s performance. Empirical investigations substantiate that our approach can proficiently and accurately automate the counting of sheep within practical farming environments. This method holds significant promise for the intelligent management of sheep farms and possesses adaptability for application across diverse livestock types, underscoring its practical utility. In essence, this study furnishes an effective and practical solution that advances precision and automation in animal husbandry. ABSTRACT: To achieve rapid and precise target counting, the quality of target detection serves as a pivotal factor. This study introduces the Sheep’s Head-Single Shot MultiBox Detector (SH-SSD) as a solution. Within the network’s backbone, the Triple Attention mechanism is incorporated to enhance the MobileNetV3 backbone, resulting in a significant reduction in network parameters and an improvement in detection speed. The network’s neck is constructed using a combination of the Spatial Pyramid Pooling module and the Triple Attention Bottleneck module. This combination enhances the extraction of semantic information and the preservation of detailed feature map information, with a slight increase in network parameters. The network’s head is established through the Decoupled Head module, optimizing the network’s prediction capabilities. Experimental findings demonstrate that the SH-SSD model attains an impressive average detection accuracy of 96.11%, effectively detecting sheep’s heads within the sample. Notably, SH-SSD exhibits enhancements across various detection metrics, accompanied by a significant reduction in model parameters. Furthermore, when combined with the DeepSort tracking algorithm, it achieves high-precision quantitative statistics. The SH-SSD model, introduced in this paper, showcases commendable performance in sheep’s head detection and offers deployment simplicity, thereby furnishing essential technical support for the advancement of intelligent animal husbandry practices.