Biometric facial identification using attention module optimized YOLOv4 for sheep

Abstract

Automatic identification of individual animals is an important step in modern farm breeding. Current best practice involves the use of RFID tags which are time-consuming for the farmer and distressing for the animal. To overcome this, we developed a sheep face recognition model through deep learning method, and realized individual identity recognition through sheep facial images. Considering the unstable recognition effect caused by the single angle of sheep facial images in the training set, we collected five kinds of sheep facial images to extract more robust facial features. Data from 67 Small-tailed Han breed sheep, aged one to two years, from two different groups of sheep, were collected over a period of one month. To recognize the identity of sheep, the YOLOv4 model was applied, and the convolutional block attention module (CBAM) was further introduced to enhance the feature extraction ability of the model. We compared the improved model with different object detection models, and the results showed that the mAP@0.5 of group1 and group2 were 91.58% and 90.61% respectively, which proved that the improved model had good recognition performance. The findings of this research could provide technical support for biometric facial recognition of sheep.