Cattle Face Recognition Using Deep Transfer Learning Techniques

In this research, a new approach has been proposed for cattle face recognition using RGB images based on deep convolutional neural network. Nowadays, biometric identification of animals is a major problem in computer vision and livestock sector. In this research, all RGB images were preprocessed to improve recognition reliability. A deep learning model was carried out using additional data augmentation methods and fine neural network tuning. The pre-trained neural networks chosen were VGGFACE and VGGFACE2. As a result, the VGGFACE2 pre-trained neural network was chosen to identify cattle faces with 97.1% accuracy.