Research on Livestock Detection Method Based on Semantic Segmentation Technology
The grassland resources in the Sanjiangyuan area are abundant,but due to overgrazing,they have been severely degraded in recent years.Cattle and sheep are important livestock for breeding in this region.Traditional detection methods have difficulties and low efficiency in counting the number of cattle and sheep,and their accuracy is easily affected by the posture of cattle and sheep and subjective human factors.Based on the above issues,this study takes the Menyuan Horse Breeding Farm and other areas as the research area,collects data on site,and uses the Yolov5 algorithm to detect cattle and sheep,meeting the requirements of high real-time detection.Compare the effects of Adam,AdamW,and SGD optimizers through experiments to improve the accuracy of recognition and counting.
the three rivers source areacounting of cattle and sheepYOLOv5 algorithmobject detection
万方数据
