Research on Re-Identification of Amur Tigers and Bengal Tigers

This study was devoted to the individual identification of different tiger subspecies, a relatively little covered research area. As one of the endangered species in the world, the tiger's population monitoring and conservation is critical. However, traditional tiger monitoring methods face numerous challenges, including difficulties in identification, inconvenient data access and high costs. In this study, we adopted a machine recognition approach using the EfficientNetB6 deep learning model as the base model and combined it with a transfer learning strategy for the purpose of recognizing individuals of different tiger subspecies. We constructed a large-scale image database containing two subspecies of tigers and labelled it in detail. Through transfer learning, we successfully applied the EfficientNetB6 model for the tiger individual recognition task, overcoming the difficulties of data scarcity and differences between subspecies. The experimental results showed that the training set accuracy of the EfficientNetB6 model was 98.0%, while the validation set accuracy was 93.0%. This study provides a novel and efficient method in the field of wildlife conservation, which is expected to improve tiger monitoring and population management. In addition, our method also offers valuable references and methodologies for individual recognition studies of other large mammals.