Related Experiment Video
Updated: Jul 24, 2025
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
A high-precision facial recognition method for small-tailed Han sheep based on an optimised Vision Transformer
Xiwen Zhang1, Chuanzhong Xuan1, Yanhua Ma2
1College of Mechanical and Electrical Engineering, Inner Mongolia Agricultural University, Inner Mongolia, Hohhot 010018, China; Inner Mongolia Engineering Research Center for Intelligent Facilities in Prataculture and Livestock Breeding, Inner Mongolia, Hohhot 010018, China.
A new sheep face recognition model, ViT-Sheep, uses Vision Transformer (ViT) technology for accurate animal identification. This artificial intelligence approach enhances livestock welfare and farm management with 97.9% accuracy.
Area of Science:
- Artificial Intelligence
- Computer Vision
- Animal Science
Background:
- Accurate individual animal identification is crucial for animal welfare and farm production efficiency.
- Radio Frequency Identification (RFID) has limitations in meeting current practical application requirements for animal identification.
- Existing animal recognition methods need improvement for precision management.
Purpose of the Study:
- To propose ViT-Sheep, a novel sheep face recognition model utilizing the Vision Transformer (ViT) architecture.
- To enhance precision animal management and improve livestock welfare through advanced identification technology.
- To compare the performance of ViT-based models against traditional Convolutional Neural Network (CNN) approaches.
Main Methods:
- Collected a dataset of 160 sheep face images for model training and evaluation.
- Developed sheep face recognition models based on both CNN and ViT architectures.
- Implemented targeted improvements for ViT, including the LayerScale module and transfer learning, to enhance feature learning and accuracy.
Main Results:
- The ViT-Sheep model achieved a recognition accuracy of 97.9% on the sheep face image dataset.
- The proposed ViT-based model demonstrated superior performance compared to CNN models.
- The study confirmed the robustness and effectiveness of ViT for sheep face recognition.
Conclusions:
- ViT-Sheep offers a highly accurate and robust solution for sheep face recognition.
- The findings support the practical application of artificial intelligence in sheep production for improved management and welfare.
- This research advances the use of deep learning in livestock identification systems.
