Automatic detection of sow posture and estrus based on convolutional neural network

Estrus detection is an essential operation in the breeding of sows, and accurate estrus immensely important to maintain productivity reproductive performance sow. However, traditional sow relies on manually back-pressure test, which time-consuming labor-intensive. This study aimed develop automatic method detect estrus. In this study, a model based optimized yolov5s algorithm was constructed four postures standing, sitting, sternum, lateral, calculated frequency posture change sows. Based this, we studied behavior sows before after The embedded convolutional block attention module into backbone network improve feature extraction capability model. addition, object box judgment used avoid interference from other region. Accelerate TensorRT platform, ensuring that graphics card can run with lower latency. result shows precision 97.1%, accuracy 94.1%. processing time single image 74.4 ms, could better meet demand production.