Amur tiger stripes: individual identification based on deep convolutional neural network.

Area of Science:

• Wildlife conservation
• Computer vision
• Deep learning applications

Background:

• Manual identification of Amur tigers (Panthera tigris altaica) is labor-intensive and limits population monitoring.
• Effective conservation strategies require scalable methods for individual tiger identification.
• Existing methods struggle with large image datasets.

Purpose of the Study:

• To develop an automated system for individual Amur tiger identification using deep convolutional neural networks.
• To assess the accuracy and efficiency of the proposed deep learning model for Amur tiger identification.
• To provide a scalable solution for Amur tiger population monitoring.

Main Methods:

• A deep convolutional neural network algorithm was developed for automatic individual identification.
• Experiments utilized 8277 images from 40 Amur tigers in Tieling Guaipo Tiger Park, China.
• The model was tested on both left and right body sides of the tigers.

Main Results:

• The automated system achieved high recognition accuracy: 90.48% for the left side and 93.5% for the right side.
• The developed network demonstrated comparable accuracy to state-of-the-art models like LeNet, ResNet34, and ZF_Net.
• The proposed method exhibited significantly shorter running times compared to other networks.

Conclusions:

• Deep convolutional neural networks offer a viable and efficient approach for automatic Amur tiger individual identification.
• This technology can significantly enhance population monitoring and conservation strategies for Amur tigers.
• The study presents a novel, scalable solution for identifying Amur tigers from large image datasets.