Open Set Sheep Face Recognition Based on Euclidean Space Metric

As the essential content of intelligent animal husbandry, identifying each livestock is only way to achieve modern and refined scientific husbandry. This paper proposes a sheep face recognition method based on European spatial metrics realizes noncontact identity by training network using image samples in natural environment. The SheepBase data set was first proposed this process, which contains 6559 images Inner Mongolia fine-wool Sunite sheep. To enhance diversity data, were data-enhanced. Secondly, solve problems more redundant information poor posture angle face, we propose detection correction (SheepFaceRepair) method. aims detect area be recognized align area. On basis, offer an open facial (SheepFaceNet) metric. incorporates biological features identity. We also tested effectiveness set. experimental results show that much higher than other methods, precision reaches 89.12%. In addition, found integrating biometrics can effectively improve network’s capacity.